­Speaker’s Notes: Am I Blue? (v2.0)

 

Brian C. Madden, Ph.D.

 

Department of Dermatology

University of Rochester

30 January 2011

 

(1) Introduction

 

No, this is not going to be an examination of my state of mind. Instead, we are going to take a look into the basis of dermatologic diagnosis by reviewing the relationships among optics, visual perception, and the anatomy and physiology of skin.

 

(This is a 2011 update to a presentation originally given in 2000.)

 

 

(2) Approach­­

 

To understand the processes that contribute to obtaining an accurate clinical diagnosis through observation, we need to appreciate the optical properties of both light and tissue and, as we shall see, a little awareness of our own perception doesn’t hurt. What establishing confidence in dermatologic diagnosis comes down to is how do we know to what degree we can accurately apprehend critical diagnostic detail?  To determine this we need to model and measure all of the components that contribute to an observation.

 

This presentation will review the four principal contributors to the appearance of skin: the physics of light, the anatomy and physiology of skin, the optical properties of skin, and the perception of skin.

 

Why are we interested in phenomena that alter observed light paths?  Light gives us an indication of the state of matter at a distance. If we can accurately model the appearance of skin, it should be possible to obtain a noninvasive estimate of the condition of tissue within the body. To accomplish this, it is necessary to have a detailed understanding of the way the pattern of backscattered and reflected light is produced. This knowledge has particular relevance here because it characterizes the substrate upon which most clinical judgments are made during a physical exam.

 

Why should we be interested in the appearance of blood vessels?  Hemoglobin is one of the two major chromophores of the skin, and besides winning me a Oscar for helping Pixar to finally create acceptable photorealistic skin for the human characters in Toy Story 4 (Of course there will be another one!), the appearance of blood and the vasculature patterns in skin provide important diagnostic information in clinical applications. These applications range from the treatment of port wine stains to the measurement of the vascular component of pigmented lesions and the diagnosis of rashes. Their proper execution require a detailed understanding to relate differences in appearance to changes in the health of the skin. It all gets a bit tricky without an operator’s manual for your visual system.

 

 

(3) Each representation of the duality of light conveys important properties

 

Characterization of the properties of light fall into two domains:

 

   In geometrical optics, light is modeled simply as rays. Rays describe the paths taken by light acting as particles that travel in a straight lines that can be altered by reflection or refraction when the light particle encounters a boundary formed by a change of material.

 

   In physical optics, light is modeled as electromagnetic waves. By representing each point of the moving wavefront as the center of a new source of a radiating wave, the superposition of the sum of these secondary waves form the future wavefront.  Characterization of light as a wave can account for much more of the diversity of the interaction between light and matter than is possible with geometrical optics (Although in this application, I’m sure Huygens would agree with Satchel Paige –  don’t look back!).

 

 

(4) Image formation

 

Through application of the myriad of geometric relations that flow from the classical properties of light and glass, come an equally diverse collection of imaging capabilities. Note, however, the simple relationships between the propagation of light and the principal points of a simple convex lens that lead to the formation of an image.

 

 

(5) Light waves

 

Light is a self-propagating wave that is composed of transversely oscillating electric and magnetic fields. These fields are orthogonal both to each other and to the direction of their propagation.

 

Light as a wave construct provides descriptions for a range of optical phenomena including interference, polarization, refraction, diffraction, absorption and scattering. The state of the oscillation of the transverse electric and magnetic fields as the light propagates determines the specifics of these wave properties and the pattern of their interaction with tissue based on its optical structure. Several of these interactions are important in the formation of the backscattered light distribution and will be covered separately in later slides.

 

 

(6) When light interacts with the surface of the skin, changes in its path will depend on the presence and distribution of these four types of events:

 

Transmission (refraction) occurs as a portion of the incident light enters the skin as a function of its polarization, has its path altered as a function of the angle of incidence and the difference in the index of refraction between air and tissue and then continues on in a new straight path through the tissue until it undergoes other event.

 

Reflection also occurs in a manner dependant on the polarization and angle of the incident light but with a path that is symmetric about the surface normal proximal to the point of the light-tissue interaction.

 

When there is a match between the oscillations of the light and tissue (specifically, between the oscillations of the transverse fields of the light wave and the induced motion in a charged particle in the medium), the interaction between a light wave and the particle ceases propagation as the light is absorbed and the particle takes on the energy of the photon.

 

On the occasions when the light-particle interaction is not a match, the interaction results in scattering – a change in the path of propagation that depends on the oscillation and orientation of the wave relative to that of the charged particle.

 

 

(7) Transmission/Refraction

 

As just indicated, light will travel in a straight line until it is subjected to an external interaction. When a light wave reaches the boundary of an optically different material that is not opaque and can support the transport of light, the direction of propagation of the light will change as a function of its angle of incidence measured from the local surface normal, the values of the respective refractive indices of the current and new materials, its polarization measured with respect to the plane of incidence, and the wavelength of the light (due to dispersion, shorter wavelengths bend a bit more). The index of refraction reflects the change in the speed of light in the associated material from that observed in a vacuum. In consequence, the wavefront appears to bend as the light crosses the boundary at an angle as a consequence of it slowing down (or speeding up) as it enters the new medium. However, at normal incidence, the light bends not at all.

 

The process of refraction is described by Snell’s Law: 

n1 sin(incident angle) = n2 sin(refracted angle)

 

Materials that support transmission can be placed along a pellucid continuum from transparent where the pattern of incident light intensities is retained to translucent where some, or all, of that pattern is lost due to scattering within the medium.

 

 

(8) Reflection is a bit more complex

 

In the optical sense, reflection refers to that portion of the light incident on a material boundary that does not cross the border into the new material but, instead, reverses direction symmetrically about the local surface normal. In this sense, the resulting overall pattern of reflected light depends on the smoothness of the surface on a scale down to the order of the wavelength of the incident light.  Realistically irregular biological surfaces, such as those formed by corneocytes, can alter the distribution of the reflected light in a complex manner with respect to its incident angle. The resulting reflectance can produce an appearance that ranges from highly specular (if the skin is wet) to matte (if the skin is dry). With Lambertian reflectance (an ideal matte surface often assumed in many applications for computational simplicity wherein the apparent brightness of an object does not change with viewing orientation; however, it is an ideal not often approached in reality), the pattern of reflected intensities from an incident beam follows a cosine distribution.

 

In a more general sense, however, reflection involves more than maintaining equal angles about the surface normal. This second interpretation bundles both the extrinsic component (surface/specular reflection) and an intrinsic component (backscatter made up from the light that crossed the boundary, was altered in path and composition by scattering and absorption and then reemerged back across the border). It is this composite (extrinsic/intrinsic) reflectance distribution that comprises the physical basis of appearance.

 

 

(9) A formal description of surface reflectance properties

 

A single diagram is not sufficient to clearly represent all the potential complexities of the 4-dimensional (elevation and azimuth of incident and reflected light) function that is the standard Bidirectional Reflectance Distribution Function (BRDF), even though characterizations of natural materials commonly possess many symmetries.

 

The four parameter function generates a measure of the radiance/irradiance ratio per solid angle in the reflection measurement direction.  The BRDF is defined for angles over a hemisphere about a point on the measurement plane by the ratio of the reflected radiance (power per unit solid angle per unit projected surface area) in the direction of the reflectance measurement to the irradiance (power per unit area at the measurement surface) from the direction of incidence. The BRDF extends to 5D with wavelength, and can be measured over time.

 

Originally formulated as a description of the pattern of surface reflectance for opaque materials, the BRDF was intended to represent the relation between the azimuth and elevation of the incident light and the angular distribution of the pattern of reflected light. This metric characterizes the surface geometry and specular reflection and contains no interaction with the subsurface material.

 

However, as was discussed in the previous slide, there are additional effects that need to be accommodated in a large number of materials (including tissue) due to light transport within the material and the reemergence of the (initially refracted and subsequently backscattered) light. The reemergence may occur at a location potentially displaced from where it was incident, a distance that depends on the local internal composition of the material, and chance. Internal passage affords this component of the light distribution to take on properties characteristic of the material due to photon absorption and wavelength-dependent scattering. Variants of the original BRDF such as the 8-dimensional Bidirectional Surface Scattering Reflectance Distribution Function (BSSRDF) have been created to accommodate the additional complexities associated with internal light transport.

 

Effects due to the size and position of the aperture used to sample the reflectance distribution (the solid angle of the output being captured, for example, by the pupil of an observer) need to be explicitly addressed in the proper interpretation of the BRDF.

 

 

(10) Photon world view

 

This SEM image approximates what a 320 nm photon sees about 100 femtoseconds before the end of its 498 second journey from the sun, just before it damages some of your DNA.

 

As is apparent in this image, the surface normal of skin can vary considerably at the scale of a micrometer or less. These nonuniformities can significantly impact skin appearance in a manner consistent with Angstrom’s 1925 internal reflection model for the darkening of wet surfaces. While portions of this image may reflect particular processing effects related to SEM, desquamation in vivo with its continual shedding of terminally differentiated keratinocytes should appear no less irregular.

 

Deviations from planarity further complicate the BRDF with the presence and consequences of erose edges, multiple interreflections, and repeated transits of light through air/tissue and tissue/air boundaries formed by loosely attached, flattened and drained keratinocyte husks that make up the surface of the stratum corneum.

 

 

(11) Back to the wave nature of light

 

There are different ways the photon’s electric field can oscillate as the light propagates. It is the orientation of the electric field that defines the polarization of light. When the oscillation maintains a fixed orientation relative to the direction of propagation, it is defined as plane, or linear, polarization.

 

Alternatively, the electric field can rotate about the direction of propagation in either a clockwise or counterclockwise direction. This pattern of oscillation is called circular polarization.

 

Along with changes in the index of refraction and in the angle of incidence, the polarization of the light determines the disposition (reflection or refraction) of the light in the interaction with tissue (as we shall see in the next slide).

 

Studies (e.g., Jacques et al., 2000) have shown that linearly polarized illumination can be used to isolate tissue detail in the epidermis and upper dermis. This ability is lost as the polarization is scrambled by collagen birefringence deeper in the dermis.

 

 

(12) Refraction/Reflection/Polarization

 

As indicated previously, the pattern of the interaction of light at a material boundary changes depending on whether the transition involves an increase or decrease in the index of refraction, on whether the polarization of the impinging light is parallel (p) or perpendicular (s) to the plane of incidence (defined by the incident light path and the surface normal of the point where the path intersects the surface), and on the angle of incidence. These relations form the basis of Fresnel Reflection.

 

For transitions into a material of higher index, any smooth surface will behave as a mirror at high oblique angles. More generally, for a given viewing pose and surface configuration, the pattern of amplitude and polarization of the light impinging on the surface (from both sides) will cause the surface to appear to the observer as a varying mosaic of summed specular and intrinsic (backscattered) light. It is important to note that the trade between reflection and refraction is a zero sum game of flux distribution that alters both the intrinsic and specular contributions to appearance; however, with respect to appearance, it is a game that also depends, in practice, on the pose of the observer and the distribution of surface facet orientations.  

 

Not only will the (often brighter) specular reflectance component, when of even modest magnitude, appreciably mask the intrinsic backscatter (returning refracted) component but, in addition, any specularity directly subtracts from the amount of light locally entering the material, decreasing the magnitude of the intrinsic component and thereby reducing its visibility even more. The sum of the two components is a complicated function of the incident light that varies across the surface of the tissue. This variation is the result of the backscattered light emerging into the air from the tissue due to the changing (over both space and time) materials and mechanisms (functions of both the anatomy and physiology) that impact the transport of light within the tissue. The complexity is further increased by the particular pattern of facets that alters the direction of propagation and magnitude of both the specular and intrinsic components.

 

Within the higher index material, looking back across the boundary, there is a critical angle beyond which there is total internal reflection of both s and p components (in comparison, the total external reflection of both s and p occurs only at as the incident angle approaches 90°, just grazing the surface).  Depending on the specific indices of refraction, this zone of zero transmission (from tissue into the air) can encompass well over half the range of incident angles. Note in the figure, the restriction of any emerging backscatter to relatively vertical angles. In the presented example, any ray of either s or p polarization more than 30° off of vertical will be reflected back into the material. Note also the proximity on both sides of the material boundary of regions of zero and of total reflection of the parallel (p) polarization component. All considered, the net effect of all this angular apportionment of light on appearance depends greatly on the relative magnitude of the s and p components. A fact that can be used to appreciable advantage in the design of diagnostic devices and procedures.

 

The particular values of Brewster’s angle and the Critical angle vary with the values of the indices of refraction. Points of total refraction are a consequence of the inability of the electromagnetic field to oscillate in the direction of propagation. In consequence and in accordance with Snell’s Law and the Law of Reflection, total refraction will occur with parallel polarized light when the reflected and refracted paths are at right angles.

                                                                                                                                        

The analytic reflectance functions above, modified for skin, can be derived from the values of the index of refraction on each side of the boundary (air = 1.0003; stratum corneum = 1.51 (Tearney et al., 1995) which results in Brewster’s angle ~56° and a Critical Angle ~42°). A range of Brewster’s and Critical Angle values have an explicit application for dermatology. Dermatologic imaging involves a range of interface configurations from simple photographic surface capture with dry skin to contact dermoscopy using a variety of interface liquids to severely reduce any specular component.

 

 

At the most basic level, the information required to characterize the resulting light distribution beyond a specific bound is unknowable (uncertain). Complexities in measurement and modeling only serve to increase this lower bound. Hence, in practice, statistical simplifications are applied to large numbers of simulated interactions and success is achieved when appropriate simplifications have been determined and the model error is reduced sufficiently to allow the prediction of patterns that are significant on a perceptual scale. Why is this important?  It is how we determine the extent to which we can both interpret and trust what we see. More importantly, it can show us what can we do to improve the process. Incorporating properties of Fresnel Reflection (and those of other light-tissue interactions) into the definition of the imaging process presents significant opportunities for the improvement of diagnostic performance.

 

 

(13) Light absorption

 

Beer’s Law:  The transmission of light through an absorptive medium displays a logarithmic relation between the residual light intensity along the path, the path length and the absorption coefficient, 

where:

A is the absorbance

            Io is the reference intensity

            I is the test intensity

            epsilon is the molar absorptivity; it compensates for pathlength (l) and concentration (c)

 

This model breaks down at high concentrations (self-screening) or when significant scattering is present.

 

We need to adjust this relation somewhat to account for the mechanisms of skin appearance and to characterize in vivo tissue and chromophore properties:

            Nonuniform concentrations of absorbers

            Complex and varying pathlength

            Scattering complications of absorption measurements

            A multiplicity of absorbers, each with different spectral and spatial distributions

 

Variations of the term:  Absorptivity is the ability of a material to absorb radiation. Absorptance is the ratio of absorbed to incident radiation. Absorption describes the process of radiative energy being taken up by particles. Absorbance is a measure of the ability of a material to absorb radiation (log10 (1/Transmittance)).

 

 

(14) Chromophores

 

Beyond contributing to the appearance of skin in the visible light range, the absorption of light well into the UV and IR ranges can initiate physiological processes and may also induce tissue damage. These diagrams cover many of the chromophores known to be present in the skin.

 

The major absorbers in the skin in the visible range are melanin and hemoglobin. Their concentrations dominate the observed variation in skin color, but there are many more absorbers of light. As mentioned in the previous slide, measurements of the in vivo chromophore distributions and effective absorbances that take all the complexities imposed by the structure of skin into account are still being worked out and await better models.

                                

Skin chromophores:

(oxy-, deoxy-) hemoglobin and other hemoglobin variants

(eu-, pheo-) melanin* and melanogenesis products (DHI, DHICA, 5-SCD)

beta-carotine (a precursor of vitamin A)

bilirubin (a product of heme catabolism)

7-DHC (leads to the formation of vitamin D3)

NADH (a UV absorber found in mitochondria)*

protoporphyrin IX (combines with ferrous iron to form heme)*

riboflavin (vitamin B2)

tyrosine (amino acid)

tryptophan (amino acid)*

keratin*

elastin*

collagen*

adipose tissue (NIR 910nm)

advanced glycation end products  (AGES)*

urocanic acid

other chromophore precursors and metabolites

DNA/RNA

various other proteins

lipids

water

 

(*) endogenous fluorophore

 

 

(15) Hemoglobin

 

Let’s take a closer look at the principal chromophore involved in the appearance of veins.

 

An examination of this figure suggests that the relative appearance of oxyhemoglobin and deoxyhemoglobin is largely due to the differences in absorption on either side of their local absorption peaks that occur in the middle wavelengths. Now compare these differences to the plot of skin reflectance. An inspection of these curves will show that changes in the region of lower absorption (higher reflectance) of oxyhemoglobin in the longer wavelengths and of deoxyhemoglobin in the shorter wavelengths are clearly insufficient to shift the appearance (dominant wavelength) of hemoglobin from red to blue with variation in oxygenation. The inability to obtain sufficiently large shifts becomes more apparent when both the wavelength sensitivity of the visual system (V_λ, a unimodal distribution of weights that peaks at 555 nm and has a full width at half amplitude of 90 nm) and the skewed absorptance function of epidermal melanin which overlays the dermal hemoglobin (thereby requiring a double pass that effectively squares its absorptivity) combine to reduce the effect of the absorption differences on appearance. That is not to say that the changes are not perceptible – consider cyanosis, for example. It is just when the changes are perceived they are commonly misapprehended and that misperception is then passed on to hue naming.

 

 

(16) Scattering

 

When an electromagnetic wave encounters a small, elastically-bound, charged particle, that particle will be set into motion by the photon’s electric field. Scattering occurs when the frequency of the electric field of the incident light does not correspond to the natural frequency of the particle. The light is scattered in a direction that is both perpendicular to the oscillation of the charged particle and to the electric field of the photon. Scattering properties depend principally on the relative size of the wavelength of the light and the dimension (cross-section) of the particle.

 

The Tyndall effect exhibits a 4^th order relation with the wavelength of the incident light in interactions with particles in suspension (colloids) in  a gas or liquid. Although Tyndall’s observations and theories set the stage for much of what was to come, he held onto the incorrect notion that pure gasses and liquids do not scatter light. He believed that scattering requires the presence of impurities in the medium.

 

Rayleigh scattering displays a similar relation between scattering and the wavelength of the incident light as observed with the Tyndall effect (4^th power relation to wavelength) for shorter wavelength light. Rayleigh scattering is an elastic interaction between the light and the particle where the energy of the light is conserved. Only the direction of its propagation is altered. Rayleigh scattering holds for the interaction between light and particles smaller than the wavelength of the incident light. This model can be extended to particles up to about a tenth of the wavelength of the light in diameter.

 

Mie scattering refers to an analytic solution of Maxwell’s equations that predicts the scattering properties of light with spheres over a larger range of sizes. It also is a form of elastic scattering. Mie extended Rayleigh’s formulation to scatterers with diameters that approach and exceed the wavelength of light. In this formulation, as the particle sizes increase, the phase function is transformed from the broad angular distribution observed in Rayleigh scattering to a greater and greater proportion of forward scattering. Improvement of this model is an ongoing area of research.

 

Thomson scattering is yet a third form of elastic scattering covering the interaction of light and very large particles. The particle is displaced very little during the period of the incoming wave. Unlike scattering by particles described above, Thomson scattering involves the scattering of photons by free electrons. There is no wavelength dependence and an equal numbers of photons are scattered forward and backward. The lack of wavelength tuning is why clouds appear achromatic.

 

 

Scattering is a major contributor to the second of the two sources of light dispersion alluded to in the discussion of reflection a few slides back. It is the one operating on a micro scale that contributes to the distribution of the backscattered light and that also contributes to making the BRDF a one-to-many angular mapping. Together with an accurate description of the distribution of absorbers, a good model of the properties of light scatter in tissue will go a long way to understanding the generation of skin appearance and what information may be noninvasively recovered from within the skin.

 

 

(17) Nature will find a way

 

So, given these optical properties and the observed distribution of chromophores and scatterers, what are some of the known sources of blue in biologic material?:

 

One impressive example of nature’s engineering prowess can be seen in the creation of structured colors. Periodic dielectric structures composed of spatial variations in refractive index in one, two or three dimensions have evolved in nature to produce a wide range of chromatic effects that would be difficult, if not impossible, to create with more common chromophore-based mechanisms. Structured colors use constructive and destructive interference to create an iridescence that is often highly saturated and often varies strongly with viewing angle. These mechanisms of color production have been independently acquired by a very wide range of species. There is evidence in the geologic record for the existence of this method of producing body coloration that goes back to the Cambrian Period, back to the time living organisms were just developing eyes, a little over 500 million years ago. The structural effects can be seen in butterfly wings, beetle cuticula, and bird plumage (esp. vibrant in hummingbirds and some ducks native to our region).

 

Structured colors in biologic material fall into three general classes: diffraction gratings (1D), multilayer reflectors (2D) and photonic crystals (3D):

 

Diffraction gratings in biologic material are formed by a series of parallel grooves with a uniform spacing on the order of the wavelength of light. With each linear element acting as a scattering locus, the grooves break up broadband illumination into its component wavelengths. The combined effect of the periodic array of scatterers is to reinforce different wavelengths (among them blue) as a function of viewing angle.

 

Multilayer reflectors are built up of alternating layers of different refractive index. As incident light penetrates this structure, it is either reflected, transmitted or absorbed at each level. Of the light that is reflected, depending on the angle of incidence, the thickness of the layer, and the wavelength of the light, it can be subject to either constructive or destructive interference. The strength of the effect can be enhanced by replicating the pattern of changes in index of refraction (instances of up to ten or more layers have been observed).

 

Finally, in 3D photonic crystals, a highly ordered lattice of structures (often spherical) with one index of reflection is imbedded in a material with a difference index of refraction. Because of the repeated pattern of the 3D lattice, the reflected light is less dependent on angle but may be subject to band gaps wherein certain wavelengths will not be propagated by the lattice. With a polycrystalline organization (multiple patches of locally precise ordering), the colors tend to be more matte and much less dependent on viewing angle. When the lattice retains a precision over large regions, the colors tend to be more saturated and iridescent, exhibiting more variation with angle.

 

 

(18) Other options

 

Sorry to say, there are no feathers that are intrinsically blue. Feathers are made from some of the same components as our skin: keratin and melanin. Only slightly less rare than blue feathers are blue chromophores among vertebrates (subphylum Vertebrata). True blue chromophores have been observed in only two species: the psychedelic fish and the mandarin fish. Very little is currently known about their chemical composition or their optical properties.

 

However, there are other sources of blue in biologic material. Water possesses an absorption band in the NIR at 760 nm which has a tail that extends down into the long wavelength visible range. It is a small peak, easily missed, right in the middle of the diagnostic notch. Unlike most other instances of photon absorption that alter the energy states of electrons, this peak is due to vibrational transitions of the nuclear motions of the water molecule. If we were tens of meters in length, and much much more transparent, our high water content could result in the same blue appearance as often seen in glacial ice (or shades of Avatar). Ice possesses an additional advantage over water in the production of blue due to wavelength-independent scattering caused by the presence of large ice crystals that form under the intense glacial pressures. The added scattering increases the average pathlength that can be achieved within a given volume. The observed blue hue is due to long wavelength absorption along the illumination path, not to short wavelength scattering away from it.

 

The final image is commonly thought to be an example of Rayleigh scattering.   When observing a clear, blue sky normal to the passage of sunlight, the sky will appear deep blue and the observed light will be polarized. As viewing is directed more towards the zenith and the density of air molecules is reduced, the source of scattering is reduced and the sky appears darker. As viewing is directed toward the horizon, a brighter, more desaturated appearance occurs. While at times this may be due to increased humidity or pollution, they are not required (Smith, 2008). The 4^th power advantage to appearance of shorter wavelengths exists for single scatter conditions. Closer to the horizon, both the density of the atmosphere and the pathlength of the light increase raising the probability of multiple scattering events which whitens the wavelength distribution reaching an observer.

 

 Although Rayleigh originally agreed with Tyndall that the short wavelength scattering required contaminants (colloids), he subsequently concluded (possibly correctly) that the molecules of the atmosphere were responsible for the blue appearance of the sky. Einstein was ultimately among those who resolved this debate in Rayleigh’s favor by demonstrating that if Tyndall’s view was correct, the color of the sky would appear appreciably more varied. However, Einstein also agreed with Smoluchowski and concluded that it was the transient density fluctuations that occur in pure fluids that induced the observed increase in short wavelength scattering. Variations in gas density effectively alter the local refractive index which, in turn, act as a short-lived particles. But with respect to the appearance of either scattering mechanism, as Einstein might say, es macht nichts. Scattering predicted from light interactions with molecules and with density fluctuations has been shown by Smith (2008) to have the same contribution to appearance. If the two mechanisms do, in fact, produce equivalent results, then we would be hard pressed to show that both are not occurring in some proportion.

 

 

(19) What’s a poor mammal to do?

 

Recently, a different form of coherent wavelength tuning has been demonstrated in primates (mandrill rump and facial skin and vervet scrota) using Fourier analysis of TEM images of 3D collagen arrays (Shawkey et al., 2009). Increasing the thickness of a layer of collagen has been shown to enhance selective scattering of blue light. This wavelength selection occurs even though the collagen bundles maintain a less than crystalline precision of ordering. With a thicker collagen layer, the blue coloration has also been found to survive the absence of an underlying layer of melanin that has been shown elsewhere to be helpful in suppressing broadband backscatter that can desaturate the coherent blue scatter.

 

A similar interaction between scattering and absorption by melanin also plays a role in the creation of blue eye color. Both the size and distribution of the radial pattern of sinuous collagen fibers observed in the stroma of the human iris and the presence and distribution (or lack) of melanin distal to that collagen determine the amount of Rayleigh backscattered blue light that is created and also whether it is subsequently absorbed prior to departing the cornea (or not). In addition, a layer of melanin proximal to the retina relative to the stroma prevents any backscatter created behind the collagen and deeper in the eye from mixing with and desaturating the blue-shifted stromal scatter.

 

In a different venue, the same contributors that interact to produce the blue appearance of veins also come together in the appearance of cyanotic neonates and, as it turns out, are subject to similar perceptual cautions. As was seen earlier on the hemoglobin slide, shifts in the dominant wavelength due to the amount of deoxyhemoglobin in veins under normal physiological conditions are small. At levels of hypoxia that are quite toxic (even fatal), shifts in the skin spectrum, while certainly apparent to observation, are insufficient to move the dominant wavelength from the red to the blue, an assertion that is supported by measured reflectance spectra.

 

 

(20) Lazuline lesions

 

The wonder of structured colors notwithstanding, the appearance of lesions in mammals is dominated by the density and distribution of chromophores present in the examined tissue. Most often in the skin, changes in appearance are dominated by concentrations of oxyhemoglobin or eumelanin, or one of their related variants, either alone, or in combination. Lesion appearance also depends on the depth of the chromophore deposition (the epidermis, the dermis or even in the subdermis), the distribution of the chromophores (encapsulated in melanosomes, melanocyte nests, or thinly dispersed as with melanin dust), and the number density of the chromophores. Often appearance will also depend on contextual anatomical and physiological variations: the surrounding skin type, alterations in scatterers due to quantitative or qualitative changes in the anatomy, and the presence of other chromophores distributed above or below, the lesion.

 

In human skin, there are a number of conditions that commonly produce lesions that possess a blue appearance – all of which result from wavelength-dependent scattering. Objective measurements of some of these lesions show the degree of their blue appearance to be, in large part, a perceptual effect. Many of the other lesion types have a measured dominant wavelength that actually is in the blue range. Most of these lesions with actual short-wavelength-dominant spectra contain concentrations of melanin, particularly melanin located in regions of the skin other than its normal distribution in the epidermis. When present in the epidermis, aggregations of melanin appear externally as shades of brown. The blue appearance caused by melanin in the dermis can be the result of an abnormal pigment transfer from cells in the epidermis (pigmentary incontinence). It can also occur through production of melanin in the dermis itself with the activity of ectopic melanocytes, or melanin can bind to exogenous pigments deposited in the dermis.

 

A recent survey (Murali et al., 2009) described these lesions as Dermal Dendritic Melanocytic Proliferations based on their shared incorporation of dendritic melanocytes in the dermis. Besides a tendency to posses various degrees of blue in their reflectance, these lesions share a number of clinical and physiological features although they come in a number of variants and display a variety of morphologies (viz., DIGM, 5^th Edition, 1999). Murali et al. divided this class of lesions into the Blue Nevus which manifests as the Common Blue Nevus and the Cellular Blue Nevus, along with several less prevalent variations, and Congenital Dermal Melanocytoses which include Mongolian Spot, Nevus of Ito and Nevus of Ota, among others.

 

While Blue Nevi actively produce melanin and have a low, but nonzero probability in some of the variants of developing into melanoma, these lesions are most commonly benign, solitary, deep-blue (with a blue dominant wavelength, not just the subjective appearance of blue) macules and papules, 1-2 mm in diameter (see upper left image). Blue Nevi commonly occur at sites where melanocytes are often present at birth – the scalp, the lumbosacral region, and the dorsa of the hands and feet where, as macules, they appear as blue discolorations of the skin. In contrast, the less common Cellular variant forms nodules 1-2 cm in diameter and can extend above the surface of the skin and will occasionally penetrate into subcutaneous tissue.

 

The blue coloration is commonly attributed to the marked aggregation of melanin in the reticular dermis that both effectively absorbs the longer wavelengths that penetrate the skin to that depth and also provides a nondesaturating, opaque backdrop for the shorter wavelengths collected above that are disproportionately backscattered and contribute to skin reflectance that survives to exit back into the air rather than being absorbed as they penetrate deeper into the dermis. The Blue Nevus is usually accompanied by a region of normal dermis just superior to the lesion (Grenz zone). Together with the heavily pigmented lesion itself, this structurally undisturbed region of dermis provides a buffer that facilitates the accumulation of short wavelength scatter. The strength of the blue coloration depends on the depth of the lesion in the dermis and on the amount of melanin in the epidermis, a potential absorber of the accumulated blue scatter.

 

The term Mongolian Spot (congenital dermal melanocytosis) is a dated expression still in common use that applies to blue-gray macules observed in a high proportion of East Asian infants. The lesion almost always located on the lumbosacral skin or on the buttocks (as in the central image in the slide). The lesions have distinct margins and are characteristically a few centimeters in diameter but, at times, may be large enough to cover the entire lower back and buttocks. The pigmented cells in a Mongolian Spot lesion tend to be less densely distributed than is observed in Blue Nevi and, perhaps in consequence, it may be why the normal structure of the dermis is not disrupted in this condition. Typically, the dermal melanocytes appear at the fetal age of 3 months and the pigmentation at the fetal age of 7 months. The natural course of this condition is for the lesion is to fade in childhood with the melanocytes going away by 2 to 10 years of age.

The nevus of Ota (oculodermal melanocytosis) is a hamartoma of dermal melanocytes that can either be congenital or acquired. It is very similar in structure to the Mongolian Spot. Four classes of the occurrence of this lesion have been described that range in increasing size from a small periocular lesion appearing unilaterally, to a lesion with extensive bilateral involvement. Most typically, nevus of Ota falls within the ophthalmic and maxillary branches of the trigeminal nerve occurring around the eyes (upper and lower eyelids, periorbital skin) and temple region, in the zygomatic area, or on the forehead, eyebrow, and nose. It can appear in any of a variety of colors (black, purple, blue-black, slate-blue, purple-brown, or brown). As can be seen in the image on the right of the slide, the pigmentation can be mottled as well. Lesions may appear as brown or brown-gray freckle-like macules on the cheeks (although in the image in this slide, the portion of the nevus on the cheek maintains the same purplish appearance as the rest of the lesion).

 

One possibility for the etiology of this condition is that the melanocytes did not complete their migration from the neural crest during embryonic development. However, this view is complicated by the observed onset of the condition both at birth and during early adolescence which may also indicate a hormonal influence. Without treatment, the nevus of Ota is permanent and does not improve with time. Although the occurrence of this lesion is much more prevalent in East Asian populations, the reported cases of melanoma associated with the nevus of Ito have been overwhelmingly found to occur in the smaller subset of Caucasian patients.

 

The nevus of Ito (nevus fuscoceruleus acromiodeltoideus) is similar to nevus of Ota. The two conditions have been found to occur in the same patient. With the nevus of Ito, the cutaneous extent follows in the distribution of the lateral supraclavicular and lateral brachial nerves. Except for the fact that the involvement is more diffuse and less mottled, the pigmentary change is the same as that observed for the nevus of Ota. An example of the more diffuse gray variant may be seen in the image in the lower left of the slide.

 

 

Objective assessment of blue lesions comes from two sources, histology and image capture. The histology provides information on the spatial distribution, density, and identity of the chromophores, while the images provide a measure of the wavelength distribution of the reflectance. Many of these conditions are relatively uncommon and a portion of the available documentation borders on the anecdotal. Even for those cases where histological data are published, rarely are the measures quantified and in a form that could be used to constrain models of light transport in tissue.

 

A cursory examination of most any clinical skin image database will illustrate problems often  associated with using dermatology images as sources of chromatic evidence. The cameras used to capture these images are rarely calibrated, the images rarely include calibration targets that would allow subsequent chromatic correction, and the results have too often been subjected to well-meaning, post-capture adjustments that attempt to bring home a scientific or educational point. Note the unusual color of the surrounding skin in the image of the Blue Nevus in the upper left. Unvetted images should be viewed with care.

 

Validation of any proposed mechanism will also depend on reconciling appearance data with the claimed generative factors to form a consistent story. For example, claims in the literature that distributions of pigment on the cheek associated with the nevus of Ota are superficial are tempered by images where the cheek appearance is quite blue – a percept requiring the presence of melanin at appreciable depths to achieve. Greater quantification should be applied in the descriptions of these lesions. The use of terms such as red-blue pigment should be replaced with a dominant wavelength, or range of wavelengths, if complete wavelength spectra cannot be obtained. By convention, the dominant wavelength of a non-spectral (purple) color is given as the spectral wavelength of the projection of the color through the white point to the intersection of the spectral locus and marked as a complement.

 

In conditions that tend to occur in darker skin, the appearance of blue may depend on the relative deposition of chromophore in the lesion versus that is present in the surrounding skin. Appearance that depends on the build-up of differential scatter of short wavelength light must also contend with the absorption of that light both by melanin in the epidermis and by hemoglobin in the dermis. Diffuse distributions of hemoglobin in the papillary plexus and in the capillary loops should provide a significant absorption opportunity that does not require local concentrations of chromophores, such as in veins and in the larger venules, to be of consequence to appearance.

 

While some general assertions about light-tissue interaction can be made with appreciable confidence (e.g., the effect scatter has on wavelength spectra and the effect chromophore absorption has in enhancing the contribution of some wavelengths to the appearance of skin while attenuating that of others), definitive resolution of what mechanisms contribute to the appearance of which conditions will require more detailed models of these lesions and better data with which to test them. It remains also to reconcile the appearance of other conditions that involve abnormal distributions of melanin but that do not produce a blue appearance (e.g., melasma, lentago maligna and actinic lentigo). As suggested above, the presence or absence of the appearance of blue in a lesion should be correlated with the presence, or absence, of the mechanisms proposed to produce blue reflectance in the skin. Obtaining a consensus on the causes of blueness in the wide variety of lesion types should go a long way to clarifying the etiology of these conditions and help to define their appropriate treatment.

 

 

(21) Models of light transport in skin: Wellman

 

Given these proposed mechanisms of blue in biologic material, are any of them consistent with what we know of the structure of skin?  To what extent have the interactions between light and tissue proposed for the creation of blue in the skin been captured in analytic models of skin?  What is known is principally due a large body of research that continues up to the present day that is based on simple physical models. These models attempt to define general properties of the light-matter interaction without attending to all of the details that would be required to provide a complete reconstruction.

 

The Anderson and Parrish (1981) paper on skin optics presented the first model that captured the major characteristics of the distribution of light within the skin. Their model described the layered organization of the skin with its spatial distribution of absorbers and scatterers, dependences on wavelength, and the contributions of these mechanisms to the initiation of and the protection from damage and disease. Considerable advances in the understanding of the skin have been made in the three decades since this paper was published, and during this time it has provided a scaffold which helped to integrate both the data that came before and the additions that followed.

 

This approach has been described as a slab model because it incorporated portions of a simple version of the Kubelka-Munk diffusion analysis that was developed in other application domains.

 

But slab is such a pejorative term …

 

 

(22) Red, white and brown

 

This diagram characterizes the basic distribution of epidermal (melanin), dermal (hemoglobin) and subdermal (adipose tissue) absorbers and scatterers.

 

Given the contribution this model has made over the years, it is deserving of a more appealing moniker.

 

Let me introduce:

 

The Neapolitan Model of Skin

 

 

(23) Statistical modeling techniques

 

Radiative transfer models need to possess two capabilities. They need to embody the interactions of light and tissue and they need to have a capacity to represent the 3D distribution of tissue detail that distinguishes each simulation. Commonly, both of these needs have been addressed using statistical approximations. Over time, these models have become more complex and are supported by more precise mathematical descriptions of the interaction of light and tissue; however, they still produce characterizations of the interactions of light and skin by extracting estimates of the gross properties of the flux distributions.

 

Much of this research paradigm was developed right here in town for the analysis of light transport through film and paper at corporations such as Eastman Kodak and Xerox.

 

 

(24) Kubelka-Munk

 

The Kubelka-Munk model is composed of two opposing fluxes. Uniform diffuse illumination is assumed along with a spatial uniformity of the dimensions and content within each layer. The resulting flux transport occurs completely along the vertical dimension. All lateral flux changes cancel each other out. Each equation contains two losses due to scattering and absorption and one gain from the scattering in the opposite direction.

 

 

(25) Monte Carlo

 

This ray diagram illustrates photon paths in an Monte Carlo simulation. Repeated simulations (often millions) vary the conditions of each trial through the use of statistical sampling of the scattering and absorption properties.

 

 

(26) Models of light transport in skin: Yale

 

Even at the dawn of the use of computer graphics in medical imaging, the utility of being able to disassemble the digital model and obtain precise measures of length, area and volume was immediately apparent. Irwin Braverman was a pioneer in the application of computer graphics in his studies of cutaneous vasculature.

 

 

(27) When voxels were scarce

 

Cherry angiomas are usually benign red papules that occur in the skin. These lesions are caused by excessive growth of capillaries and are increasingly common with age. As a rule, they are quite small initially but, over time, can expand by a factor of 10 to 100.

 

The wonderful truth about this model (on the left) is that it was made from photographs of serial sections and was physically constructed out of paper, string and glue – a real 3D model.

 

 

(28) Models of light transport in skin: UT Austin and UC Irvine

 

Unfortunately for simulation design, all disease processes do not manifest themselves in the shape of a slab with uniform properties. The investigation of vascular lesions that were non-responsive to laser treatment required a more accurate representation of the tissue distribution so that the therapeutic application could be improved.

 

                                                                                                    

(29) Port wine stain

 

In this study, a Monte Carlo simulation was used to determine the laser energy absorption throughout an arbitrary port wine stain pattern.

 

Port wine stains result from a superabundance of blood vessels that produce a disfiguring vascular lesion on the skin. The lesion can be removed using a pulsed laser of the appropriate wavelength (e.g., 585 nm) proximal to a hemoglobin absorption peak. Unlike with the Kubelka-Munk simulations, the correct estimation of the laser energy deposition requires accommodating an arbitrary vascular distribution. The simulation operates on a 3D matrix of tissue properties that are used to control the simulated scattering and absorption of the incident laser light.

 

The simulation matrix with 2μ voxels that are populated with values obtained from a series of 6μ histology slides of a port wine stain is the domain through which the simulated photons propagate. The simulation is controlled by four optical properties: an absorption coefficient, a scattering coefficient, an anisotropy factor, and the index of refraction. These optical properties determine the probabilities of absorption and of scattering as well as the degree to which the path of the scattered photon will likely be changed. In the simulation, note the high intensities just below the surface. Intensities 25% or more greater than the incident flux levels can be observed, largely due to internal Fresnel Reflection at the tissue-air boundary.

 

 

(30) Models of light transport in skin: Hokkaido

 

Going a step beyond what can be afforded by standard histological sections, these next movies were assembled using a series of 30 sections brought into Photoshop and stitched together to form a 3D model (using a later technology, but similar in concept to Braverman’s physical model building technique).

 

This research was done by Tetsuri Matsumura (and a group at the Hokkaido University School of Medicine) who otherwise studies Birbeck granules in Langerhans cells. The 3D models are of normal epidermis, a BCC and an SK – simple, elegant, illuminating. By viewing these models, it is possible to identify a number of 2D histological sampling artifacts. For example, in normal epidermis, epidermal ‘pegs’ of varying widths were shown to be slices of a continuous honeycomb pattern of ridges that project into the dermis and surround the capillary loops.

 

Keep that 3D reconstruction of the rete ridges in mind. It will be of interest later on.

 

 

(31) What observations, evidence, or dogmata do we need to reconcile (or lay to rest)?

 

These models describe mechanisms in the skin that govern the transformation of illumination into appearance. The light we see as skin reflectance has run a double gauntlet. After entering the skin, it had to pass through tissue that altered its amplitude, spectral distribution and direction of propagation. However deep the light penetrates, the portion that becomes part of skin reflectance has to make its way back through similar modifications to once again traverse the stratum corneum-air boundary and exit the skin.

 

What mechanisms have been the suggested as sources for the appearance of blue in human skin?  They need to be sorted out – fact from fiction. Which of the proposed mechanisms induce the greatest effect?  Which ones are insignificant?  What claims are simply not true?  (The principal mechanisms that control light distribution in human skin are absorption and scattering, but there are also anatomical, physiological and perceptual influences. Caution should be extended in the interpretation of terms such as ‘has a blue cast’ or ‘has a blue-red hue’. Isn’t this what we are trying to ascertain: is the appearance of blood in veins red or blue?  Measure twice, hypothesize once.)

 

Iridescence such as occurs in bird feathers and butterfly wings as a source of blue in biologic material. (Iridescence is not a factor in human skin appearance. The required fine crystalline structures or linear, uniformly-spaced bundles of collagen needed to support this form of coherent interaction have not been observed there.)

 

Collagen differentially reflects short wavelengths which are superimposed on additional light from the deeper veins where the longer wavelengths are absorbed, leading to a subtractive color mixing in which scattering plays no part. (Findlay, 1970).  (The structures proposed by Findlay that are required to produce selective transmission and reflection are not a factor in human skin appearance. There is even less evidence for a subtractive color mixture mechanism. While human skin is not host to the periodic, crystalline structures that alter appearance with viewing angle, there are instances of quasi-ordered arrays of parallel dermal collagen fibers found in the blue markings on mandrills and some other primates that can produce an appearance of blue that does not appreciably vary with viewing angle as a result of scattering (Prum and Torres, 2004). However, neither the required thick, ordered layers of parallel collagen nor the absence of melanin superior to the collagen layer has been observed in human skin.

 

Blue eye color is another example of short wavelength contributions to the appearance of biologic material. In this case, however, the blue color is thought to be generated by Rayleigh scattering from the undulating radial collagen bundles in the iris and is enabled by the presence of ample melanin beneath the collagen and not above, a pattern in contrast to the organization observed in human skin. However, depending on the specific spectral weightings and chromophore densities involved, it may be possible that the relative lightening that attends scar maturation may be due in part to a desaturation caused by the deposition of parallel bundles of collagen during repair. In consequence, and distinct from the reflectance properties of non-damaged skin, the scar tissue may be able to produce at least a modicum of coherent scattering skewed toward the shorter wavelengths.)

 

There is greater long wavelength absorption by deoxyhemoglobin between 600 and 700 nm than by oxyhemoglobin. The physical spectral differences between oxyhemoglobin and deoxyhemoglobin when adjusted for human photopic sensitivity could result in the required hue shift into the blue range. (Along with the physical absorption and scattering spectra, the evaluation by human observers of the contribution of absorption or scattering across wavelengths in the skin requires photopic weighting which peaks in the middle of the visible range, however, the difference due to changes in oxygenation are small relative to the total skin reflectance spectrum and the dominant wavelength is still very much red in appearance under normal physiological conditions.)

 

Deoxygenated blood is blue, e.g., assertions in DIGM, Fitzpatrick, 4^th Ed., and many others in the medical literature to this day. (A deeply ingrained and erroneous cultural illusion that goes back at least as far as the struggle between the Moors and the Castilian Spanish where the blue appearance of veins was used as evidence of racial purity.)

 

Ex vivo blood always appears red to us because of rapid oxygenation from contact with the air. (Not true. Is the sample blue when you draw blood?  Why not?)

 

Rayleigh scattering increases as the 4^th power as the wavelength decreases - by a factor of 10 or so between 700 and 400 nm - so the backscatter of is disproportionately larger at short wavelengths and can shift the chromatic appearance of skin. (True, but not by enough. The in vivo contribution of short wavelength scatter to appearance is mitigated by the presence of other chromophores and scatterers and the weighting by photopic visual sensitivity, all of which contribute to a much greater amplitude of long wavelength light that is too much for this scattering effect to move the dominant wavelength anywhere close to the short wavelength end of the visible range.)

 

The hemoglobin in the arteries isn’t more visible because arteries are under greater internal pressure, are thinner in diameter but have thicker, more optically dense walls and are situated deeper in the subcutaneous adipose tissue than veins so as to be better protected from injury. (All true but that last bit. Every actual 3D reconstruction of the cutaneous vasculature I have found in the published literature indicate that the arterial and venous systems possess very similar distribution patterns.)

 

Why doesn’t the vasculature in the upper dermis present a blue appearance as well?  (These smaller vessels don’t contain enough concentrated hemoglobin to reduce the brightness locally. They also aren’t deep enough to become the backdrop of sufficient amounts of shorter wavelength scattering to desaturate whatever local backscatter from such deep veins makes it back to the skin surface.)

 

Does the skin possess an environment conducive to color contrast?   (Yes. This chromatic induction mechanism is tuned for a range of object sizes that possess gradual boundaries and less bright and saturated reflectance distributions as compared to a larger, brighter, more saturated surround.)

 

 

(32) What physical evidence do we have in hand?

 

Diffuse vein boundaries, short wavelength scatter, large vein diameters, thin vein walls, and hemoglobin absorption – the question is whether or not this cluster of fairly small effects is sufficient to induce the claimed large perceptual shift in hue.

 

 

(33) What do we as observers bring to the table?

 

Vision does not come with an operator’s manual. This is unfortunate since major shifts in appearance often occur because of changes in the perceiver, not in the perceived. What is perceived is affected by basic properties of the HVS such as a differential sensitivity across wavelengths and to different spatial luminance patterns. Perception is also altered by the inability of the HVS to respond veridically to the extensive range of environmental input. Over the last two decades, a more complex model of adaptation in the HVS to environmental changes has been developed that better describes how our perceptions appear to remain robust in the presence of extensive variations in stimulation.

 

Then there are adaptations that induce nonveridical perceptions that help stabilize appearance in the face of environmental change that may not have entirely beneficial consequences in the diagnostic arena. For example, the perception of hue is an important diagnostic dimension. Are hue judgments constant modulo other light properties?  Consider the mechanisms that turn dark orange into brown.

 

Also thrown into the mix are expectations, such as the common, almost universal, belief that the blood in veins is blue, that lead observers to both discount some and embellish other perceptual evidence.

 

 

(34) Luminance and chromatic illusions

 

Assessing complex skin patterns of luminance and chromaticity is a difficult task. At times it is made more difficult because we perceive changes in appearance that do not reflect material changes in the tissue, but rather are the consequence of adaptation or other nonlinearities in our visual system. The changes in many of these circumstances often involve attempts by our visual system to maintain perceptual constancy in a changing viewing environment that often contains a range of stimulation that is greater than can be represented in the neural response. Largely through uncontrolled observation conditions, these adjustments often do little but increase the variability of diagnostic decisions because the perceptions do not reflect the underlying disease processes in any reliable way.

 

This slide presents a few examples that should help you appreciate how well you’re in touch with your visual system.

 

The Bartleson-Breneman effect:  This figure is an excellent touchstone for the display of the inner workings of the brightness contrast effect. The luminance of the squares in each row is constant. All the columns possess identical patterns of luminance decrements. Only the oblique background gradient varies diagonally across the grid. Note that the apparent range of brightness in the right column is greater than for the left. Historically, photographs often have been mounted on a white mat because they are similarly enhanced. Observe the perceived differences in the two (identical) diagonals. The only difference is that one is normal and the other is parallel to the background gradient.

 

Ted Adelson’s Shadow-Checkerboard illusion:  Our visual system contains mechanisms that attempt to extract surface contrast uniformly in the presence of varying environmental conditions. The squares marked A and B have the same luminance. Most of you have been using your visual system for a few years now. You’d think you would be in better touch with it than this!  There are several other examples in this checkerboard where context turns physical equivalence into perceptual disparity.

 

Luminance gradients in the blue wedges induce perceived changes in the gray background even though the luminance of the gray background is entirely uniform – a simultaneous contrast effect. The introduction of color increases segmentation of the image in ways that make the interactions more difficult to consciously recognize.

 

Another simple simultaneous contrast effect can be seen with a uniform bar superimposed on a luminance gradient – an illusion we will return to later on. Outside carefully constructed views in a vision research lab, this effect is probably present to some degree in almost everything you have ever seen.

 

 

(35) Illusions with common skin tones

 

These examples show what induction effects can be obtained with standard skin tones (there are no colors that can’t be found on my right wrist). Two of these squares are identical. They are uniform patches, identical to one another in all respects. The other two differ and are non-uniform.

 

The two squares on the right are uniform patches of skin color. The two on the left have luminance gradients: top (90, 78) and bottom (100, 68). Even a moderate linear luminance gradient (90 to 78) can be difficult to discern.

 

 

(36) Illusions with common skin tones and edges

 

These examples show what induction effects can be obtained with standard skin tones and the addition of narrow luminance dipoles referred to as Cornsweet edges. Local changes in contrast induce wide-spread perceptual changes because we are more sensitive to the sharp transition in one direction than to a slower change of equivalent magnitude in the opposite direction.

 

On the left, the superposition of a Cornsweet edge in the middle of two continuous gradients alters the perception of gradual change into step functions. On the right, opposite polarity Cornsweet edges can induce a step function in a region of uniform luminance. Because of adaptation effects, the illusion is often more compelling if you fixate on one of the dipoles while making brightness judgments.

 

What is, in fact, uniform can appear to be nonuniform. What is, in fact, nonuniform can appear uniform. An inducing edge could be a hair shaft or a superimposed specularity. More of concern are the effects induced by exaggerated edge sharpening (unsharp masking) and contrast enhancement that are used in some cameras to improve image appearance. In some circumstances, the changes due to these aesthetic filters can approximate the effects of Cornsweet edges.

 

 

(37) What’s in an image?

 

One lesson to learn from this presentation is that care should be taken when making observations to ensure that viewing conditions are appropriate for the perceptual (or diagnostic) task at hand.

 

For example, while this blue background provides an aesthetically pleasing backdrop to the skin tones in the wrist image and has some historical precedent from long use in the anatomy literature, it is perhaps not the best environment against which to make a judgment of skin coloration.

 

 

(38) You sure?

 

In this macro image, a slightly rotated 2 cm vascular segment taken from the previously displayed wrist image is displayed against a gray surround. Measurements show that there are, at best, only very small changes in hue. It is clear that these measured changes in hue at most a couple of degrees around the color circle are totally insufficient to account for the proposed change in color perception in the range of 120°.

 

All the skin hues in this image are in a small cluster separated from the blues by an appreciable distance around the color circle. A cursory examination of the segmented pixel regions will show that the veins are visibly distinct largely due to decreases in luminance and in saturation.

 

The magnitude of the observed perceptual effects of the veins is reduced in this enlarged segment taken from the previous image because of the spatial frequency sensitivity of the induction process – size often matters.

 

 

(39) Open bar

 

Consider the bar which has the approximate average HSB values of the skin in the prior macro image (22, 37, 84) that is positioned near the bottom of the skin-toned background. The contrast of this bar is in a range quite common for skin patterns.

 

The background is actually a linear brightness gradient (90-78) with hue and saturation constant, the same as one in the upper left of a figure containing gradients and uniform patches we saw a few slides ago.

 

The bar at the top is, indeed, identical to the one at the bottom.

 

 

(40) Crossing the bar

 

As is the bar now connecting them.

 

All the perceptual differences along the bars are induced by a subtle, vertical luminance gradient. Such gradients are easily produced by a nonuniform distribution of chromophores in the skin or by minor variations in the illumination of the exam room. A steeper gradient and more extreme bar settings could have been employed while still using values that are possible within the observed gamut of in vivo skin tones. The enhanced selection would increase the magnitude of the induction effects. However, the current values were selected to be more moderate and are situated well within the range of the observed skin tones – well within the norm of the distribution for living dermal tissue you would run into in most any patient. This sort of induction happens all the time.

 

 

(41) Gray is the new blue

 

Note the changes that occur in the induced color and brightness differences, and in the time course of those changes as I alternate the gradient and flat gray backgrounds. The induction effects aren’t quite instantaneous. Note also the ability of the skin colored gradient to induce a blue chromatic aftereffect on the gray background.

 

The gray background was selected to have a close match to the bars in brightness (71), or (180, 180, 180) in RGB. The difference in the matched brightness values between the background (71) and the bars (84) probably has much to do with the need for photopic weighting, and some display calibration would probably help here as well.

 

 

(42) My wrist

 

Three 1 by 10 mm samples were obtained from the wrist image, one along a vein and two others just above and below. In this demonstration, no effort was made to rigorously control the uniformity or specularity of the illumination or to calibrate the capture process other than to use a standard clinical camera configuration. It is just an image I captured for this talk over 10 years ago.

 

However, the reflectance difference in the vein sample is real. The perceived and objective differences are just not due to a change in the hue of the stimulus. The significant differences are in the saturation and brightness values.

 

It might be time to recalibrate, not the camera, but the faith placed in an adaptive, nonradiometric visual system.

 

 

(43) My goodness!

 

The square in the upper left reflects the average HSB values in the macro wrist image. The other squares present the differences between those average values and the average saturation and brightness values for the vein sample, alone and in combination.

 

It is interesting to observe that the magnitude of the perceptual effect due to the decrement in brightness in the vein sample appears to be larger than the effect due to the decrement in saturation (the values come directly from the observations; no attempt was made to select or to adjust them). The greater effect observed for decrements is consistent with commonly observed simultaneous contrast results. While the decrease in saturation alone would appear to have a lightening effect that might be expected to act in opposition to the brightness decrement, examination of the differences in the right column (brightness difference with and without the saturation difference) seem to support a more complex interpretation.

 

It is true that the brightness decrement alone does appear to be darker. It also appears to not possess much of a shift in perceived hue. Apparently, it is the admixture of the two decrements in the vein image that induce a blue-gray shift in appearance that is attributed to ‘blue blood’. This perception is also aided by an ingrained and pervasive belief with respect to a blue appearance of the veins that is held by patient and clinician alike.

 

A key contributor here is that scattering grows as much as the 4^th power of the wavelength of the light and while this contribution to the skin’s reflectance is not sufficient to shift the perceived hue from the reds to the blues, it is apparently enough to sufficiently desaturate the veins’ appearance so that, coupled with a slight decrease in brightness and a blurring of the boundary of the veins, it becomes a good candidate for the color contrast effect.

 

In a recent study on port wine stains (Widdowson et al., 2010), perceived chromatic change in veins was investigated as a function of their depth. This measure is of interest since clearance of these defects with laser treatment depends largely on the depth of the veins. Establishing a noninvasive measure of vein depth would allow assessment of likely outcomes to be accomplished prior to treatment based on skin imaging.

 

In their study, the brightness (HSV color space) of the phantom was found to remain constant as its depth increased. Isn’t that somewhat unexpected?  Why wouldn’t V change with increasing depth as more light from the brighter surround tissue gets diffused over the regions of reduced vein backscatter?  Shouldn’t scatter move the observed V measures toward the mean luminance of the surrounding skin?  In a study of the use of fingerpad vasculature pattern as a recognition biometric (Lee and Park, 2009), the sharpness of blood vessel boundaries observed from the surface became increasingly blurred with vessel depth. This alternate measure of diffusion in the backscatter lends support to a model where both the diameter of the vessels and the amount of superimposed surround diffusion increase with vessel depth, possibly resulting in a rough constancy of reflectance over the vessels.

 

Saturation was found to decrease from 75% to 5% for a 5 mm change in depth.

Hue was found to change from -7 deg to -38 deg, about 25% of the 120 deg distance from red to blue on the HSV color circle. The common depth of veins in the skin is more likely to incur a hue shift of half the peak value found in the simulations. The values in the simulations are also in contrast to my measurements of skin hue that were about +20 deg in the red-orange range, almost completely opposite to blue in the HSV color circle.  The hues in my data cluster within a couple of degrees both for veins and surrounding skin. Also, what about the findings by Lakmaker et al. (1993) that find minimal contribution to appearance from vessels much below 1 mm?  These findings call into question the appropriateness of the phantom as a surrogate for in vivo color estimation but still may be cobbled together with the other findings to make a consistent story of diameter, depth, diffusion and density.

 

 

(44) Funny, they don’t look the same

 

The bars on the left possess the approximate average HSB values (22, 37, 84) of the wrist macro image, as before. The bars on the right possess the approximate average HSB values (22, 33, 79) of the vein sample.

 

The background is a top to bottom, linear brightness gradient (90-78), hue and saturation are constant, all as before.  Again, the bars at the top are identical to the corresponding bars below.

 

Please note your mind at work here (this figure is a caution to never leave it unattended or without conscious supervision). Consider while you scan the slide that the hue value (22) is the same for every pixel across the entire slide, both bars and gradient.

 

At some point in the distant past, incorporating these perceptual adaptations into the human visual system must have increased the survival potential of one of your ancestors. Nonetheless, it should be recognized that no matter how longstanding this advantage has been to your genetic lineage, it may not carry over to the exam room later this morning.

 

 

(45) Faux blue filter

 

The previous slide demonstrated some of the perceptual changes that can occur while holding hue constant. This knowledge can be applied to skin images to locate changes in appearance that are not associated with a hue differences. Of import here is that when applied to the wrist image, it is possible to isolate the distribution of veins without using hue.

 

The top image is a thumbnail of the wrist macro image. The middle row contains scaled, pseudocolor representations of the hue, saturation and brightness values of that same RGB wrist image. The small changes in hue can be largely attributed to specularities and shadows (due to variations in contributions from room and flash illuminants) present in a candid capture environment that was not very well controlled. Patterns that correspond well to the location of veins can be seen in the saturation and brightness images. A simple algorithm based on a pointwise product of saturation and brightness values yielded an even better correspondence to the veins than either property alone. Thresholding this computed image provided a template for the location of the perception of blue skin color.

 

 

(46) Unresolved issues

 

There are several proposed mechanisms and interactions whose integration could potentially contribute to a better model of the appearance of veins:

 

To what extent do physiological factors such as hematocrit and oxygenation alter vein appearance in vivo?  What conditions and controls should be in place to appropriately measure the spectrum of hemoglobin in vivo?

 

Blood measurements not obtained under pressure and flow may be corrupted by sedimentation and clustering. Clearly, dense packing of chromophores can alter the effect they have on skin reflectance and, just as clearly, the oxygenation of blood can perceptibly alter skin color as well when there are large changes, as in cyanosis.

 

To what extent do dermatoglyphic patterns and surface hydration alter the appearance of veins at the surface  - especially when they act as visible striations that may break up diffuse gradients associated with deeper structures?

 

Do Fresnel Reflection and the extent of internal reflection of backscattered light at the skin/air boundary (Critical Angle) have a role to play in the modeling of the external appearance of veins?  The question that needs to be asked when contemplating the contribution of a given effect is how much of its alteration of the pattern of light approaches the surface of the skin within 42° of the surface normal and how much does that portion of the backscattered light differ from the portion that is not within 42°?  How directional is the appearance of ‘blue’ veins when viewed at angles appreciably away from the surface normal?  What complexities are added to the already spatially complex epidermal-dermal boundary with the inclusion of an increase in the index of refraction going into the dermis?

 

The relation between saturation, brightness and the shift in perceived hue in the vein reflectance was uncovered ten years ago in preparing a simple demo for this talk.  Will the wrist sample findings hold up with more substantial sampling of vein reflectance characteristics and with better control over lighting and specularity?  In particular, is the assertion that both measured saturation and brightness values decrease by a small amount over the veins while hue changes are negligible supported by a larger, better controlled sampling of vein reflectance?  What are the width and the gradient parameters of these perceptual effects with respect to the actual dimension of the veins?  How much do these measures change with the depth of the veins?  Also, it should be helpful to fold in the many additional details (above) that have significant contributions to vasculature appearance, a good portion of which have been reported over the last decade due to an increased interest in noninvasive blood assessment. These findings could be incorporated in a device or procedure to better isolate the contributions to the surface reflectance of veins (and arteries) that better represent the underlying anatomical, physiological and perceptual contributions.

 

A more accurate measure of the induced hue effect might be obtained if a more spatially detailed method of segmentation of the surface representation produced by the underlying vasculature were incorporated. This addition could improve the strength of the color contrast effect by softening the segmentation boundaries as the luminance of the surrounding skin is varied. It also would be helpful to obtain independent estimates of blood oxygenation and hematocrit, vessel depth and wall thickness. Besides ultrasound, in vivo confocal microscopy, and OCT, there are a number of established and emerging blood measurement devices that could be put to good use here.

 

Is Land’s retinex theory the best candidate to model the color contrast effects observed under these conditions?  How does the New Standard Model account for color contrast effects? 

 

Representations of the structure of the dermal vasculature in several articles on light distribution in skin appear to be at odds with 3D reconstruction appearing in a different part of the literature on the development of new surgical techniques. Many of the former data are based on research that goes back to the 19^th century. Caution should be exerted here in that this was a period where the scientific criteria for data collection waxed interpretive along the lines of Schiaparelli’s canali. The subcutaneous veins are connected to the reticular vascular plexus deep within the dermis that is itself connected through descending venules that serve mid-dermal structures such as sebaceous glands and hair follicles through shunts from the ascending arterioles. Care should be taken to include site-specific phenomena such as in the scalp where deep penetration of hair follicles drag the dermis boundary and the associated arteriovenous anastamoses into what is normally the adipose tissue/subcutis level (just as they often tend to drag epidermal structures down into the dermis on the other end of the follicle). These venules originate from the papillary plexus that connects to and drains the capillary loops just below the dermoepidermal junction. While we refer to the reticular and papillary plexi as distinct, discrete distributions, several of the vascular reconstructions tend to support the distribution of vessels in the dermis as more of a gradient, denser at the upper and lower boundaries and less concentrated in the middle.

                                                                                                                      

Does the chromatic illusion of blue veins involve only the largest subcutaneous veins, or are elements of the dermal reticular plexus large (50µ) and deep enough to induce the perceptual effect as well? 

 

Why doesn’t short wavelength absorption by melanin in the epidermis extinguish any effect by short wavelength scatter in the dermis?  Isn’t melanin anterior to the stroma of the iris sufficient to turn potentially blue eyes brown?  What is the concentration of melanin in the anterior iris relative to that in the epidermis?  On the other hand, if it isn’t short wavelength scatter above the veins that desaturates that reflectance, then what does?

 

What scattering mechanisms need to be in play in the light paths that reach the veins to induce the required perceptual changes?  How much of the scatter is due to collagen in the dermis?  What are the other sources of scatter and how do they vary with wavelength?   The contribution of scattering to appearance depends on the distribution of light from other factors in the environment on which the scattering is superimposed. Perhaps a metric similar to optical albedo is needed to quantify the impact the available scattering has on reflected light attenuated by absorption.

 

Does the more diffuse distribution of hemoglobin in the smaller arterioles, capillaries and venules in the papillary dermis (much closer to the surface) appreciably increase the absorption for this amount of chromophore as compared to being concentrated in the larger arteries and veins deeper in the skin (c.f., the melanin dust effect increasing absorption due to minimal self-occlusion).

 

 

(47) Pay close attention

 

No, it isn’t. That slide presented the Cornsweet dipoles. This is the same slide as 10 slides ago. Now, I need for you to pay better attention. We are coming up on the take-home message.

 

I would like for you to look closely at this slide and then at the series of 11 slides that follow. In this sequence, some parts of the image will change, some will not. Appreciating the difference will go a long way toward enabling you to understand how the strengths and weaknesses of your visual system can impact diagnosis.

 

 

(48-58) True blue

 

[First run through the sequence one time without comment.]

 

In this series, the pixels in two vein branches in the wrist image were isolated. All the other (RGB) pixel values in the image were halved, reduced by 1/4, 1/8, 1/16, 1/50, remained unmodified as captured, were increased by 1/50, 1/16, 1/8, 1/4 and were increased by half. The original image pixel values were scaled back 75% to reduce the amount of clipping at the highest brightness increment. (Better assessments of image quality differences should be obtained in subsequent versions of this talk through the use of  the increased dynamic range available in RAW format images.)

 

As the relative brightness of the surrounding skin is varied, the pixel values of the isolated vein segments do not change – not in chromaticity, not in saturation, and not in luminance.

 

 

(59) Summary and Conclusions

 

The principal arteries and veins that supply the skin run below the dermis cushioned by a layer of adipose tissue. This blood supply feeds a network of smaller vessels in the lower dermis, the reticular plexus, which in turn, through ascending arterioles and descending venules, supply the upper (papillary) dermis and the matrix of capillary loops that arise out of it.  The deep arteries are smaller in diameter than the corresponding veins and have thicker walls to sustain the greater pressure. In consequence, the arterial blood is less visible because of the minimal backscatter from their contents reaches the surface of the skin. While this description of the cutaneous vasculature generally holds throughout the skin, care should be taken to note often appreciable deviations in site-specific anatomy, such as those associated with scalp follicles or nail folds.

 

The veins have greater diameters and thinner walls. They are under less pressure and act as a reservoir for the circulatory system. The small increase in absorption by venous blood that makes its way to the surface is sufficient to induce small decreases in perceived brightness and the depth of the veins also allows an accumulation of shorter wavelength scatter to be superimposed on that of the darker veins thereby inducing a decrease in the saturation over the diffuse surface representation of these vessels. These differences, in turn, are sufficient, given the fuzzy spatial distribution, to incur a perceived change in hue complementary to the brighter, more saturated, red-hued, surrounding skin through the color contrast effect.

 

While the brightness difference between the appearance of the veins and the surrounding skin is enough to be amplified by simultaneous brightness contrast (which shifts color discrimination into a darker, muddier regime where chromatic judgments are not as sensitive), it is the superimposed increase in backscatter (the Rayleigh scatter is dominated by short wavelengths but involves a range of wavelengths throughout the visible wavelengths whose contribution to appearance is weighted by a unimodal function with a peak at 555 nm where human sensitivity is the greatest; the effect of fourth power scattering function is mitigated by the decreasing visual sensitivity to shorter wavelengths; this interaction may account for the empirically observed smaller (e.g., 1.7) exponents for the relation between wavelength and scattering) that desaturates the dominant longer wavelengths of the veins and makes their appearance more susceptible to an induced change in color perception.

 

Although individually not large in magnitude, these very real effects that are potentiated by the common belief that veins are blue. Although blue tends to be darker in appearance than some other colors, it is a mistake to turn this perceptual correlation into a causal relation and to believe that because a region is darker, it is blue. Perhaps a start at a remedy for this mass cultural illusion would be for medical artists to cease portraying veins in blue while the arteries are shown in red.

 

 

This is a good time for you to ask yourself what other preconceptions you might be bringing with you into the diagnostic arena that alter what you see, or don’t see?

 

 

(60) Consider this when next you examine a patient:

 

Is all that you perceive actually there?

 

Is all that is actually there perceived by you?

 

To what extent do you need to be aware of the viewing conditions during a physical examination or when capturing diagnostic images?

 

Can you safely base clinical judgments on adaptive mechanisms developed to help your progenitors recognize bad fruit on both sunny and cloudy days, or to detect a predator in the tall grass after the sun goes down?

 

 

So, the question remains, how should you interpret what you see?

 

Consider instead, when diagnosing suspect lesions, the potential benefit of imaging utilities that allow you to accurately visualize the distribution of chromophores in all types of skin, free from perceptual distortions and environmental contamination, and supplemented with the evidence provided by 3D models of the epidermis and dermis noninvasively captured straight from the patient in vivo. How much more could you do for your patients?

 

[Now would be a good time to pass the hat. Brian needs a(nother) new camera.]

 

 

(61) Thank you!

 

At times I'm bluish, … or at least apparently so.

 

References

 

Agache, P. G. and Dupond, A.-S. (1994) Recent advances in non-invasive assessment of human skin blood flow. Acta Derm. Venereol. (Stockh), Suppl. 185:47-51.

 

Anderson, R. R. and Parrish, J. A. (1981) The Optics of Skin. J. Invest. Dermatol., 77:13-19.

 

Angstrom, A. (1925) The albedo of various surfaces of ground. Geografiska Annaler, 7:323-342.

 

Bagnara, J. T., Fernandez, P. J. and Fujii, R. (2007) On the blue coloration of vertebrates. Pigment Cell Res., 20:14-26.

 

Barsky, S. H., Rosen, S., Geer, D. E. and Noe, J. M. (1980) J. Invest. Dermatol.,

 

Borovoi, A. G., Naats, E. I. and Oppel, U. G. (1998) Scattering of light by red blood cells. SPIE, 3194:295-304.

 

Braverman, I. (1983) The role of blood vessels and lymphatics in cutaneous inflammatory processes: an overview. Brit. J. Dermatol., 109, Suppl. 25:89-98.

 

Braverman, I. (1983) Ultrastructure and three-dimensional reconstruction of several macular and papular telangiectases. J. Invest. Dermatol., 81:489-497.

 

Braverman, I. (1989) Ultrastructure and organization of the cutaneous microvasculature in normal and pathologic states. J. Invest. Dermatol., 93(Suppl.):2S-9S.

 

Braverman, I. (2000) The cutaneous microcirculation. J. Invest. Dermatol. Symp. Proc., 5:3-9.

 

Breunig, H. G., Studier, S. and Konig, K. (2010) Multiphoton excitation characteristics of cellular fluorophores of human skin in vivo. Opt. Exp., 18:7857-7871.

 

Cornsweet, T. N. (2003) A simple model for filling-in, contrast, contrast constancy and assimilation. Visual Pathways Inc., Technical Paper, www.visualpathways.com/fillin.html (accessed 2/16/2010).

 

Cui, W. and Ostrander, L. E. (1990) In vivo reflectance of blood and tissue as a function of light wavelength. IEEE Trans. Biomed. Eng., 37:632-639.

 

Dorsey, C. S. and Montgomery, H. (1954) Blue nevus and its distinction from Mongolian spot and the nevus of Ota. J. Invest. Dermatol., 22:225-236.

 

Eaton, W. A., Hanson, L. K., Stephens, P. J., Sutherland, J. C. and Dunn, J. B. R. (1978) Optical spectra of oxy- and deoxyhemoglobin. J. Am. chem. Soc., 100:4991-5003.

 

Eskew, R. T. Jr. (2009) Higher order cortical mechanisms: A critical review. Vision Research, 49:2686-2704.

 

Findlay, G. H. (1970) Blue skin. Br. J. Dermatol., 83:127-134.

 

Friebel, M., Helfmann, J., Muller, G. and Meinke, M. (2007) Influence of shear rate on the optical properties of human blood in the spectral range 250 to 1100 nm. J. Biomed. Opt., 12, 054005.

 

Friebel, M., Helfmann, J., Netz, U. and Meinke, M. (2009) Influence of oxygen saturation on the optical scattering properties of human red blood cells in the spectral range of 250 to 2000 nm. J. Biomed. Opt., 14, 034001.

 

Friebel, M., Roggan, A., Muller, G. and Meinke, M. (2006) Determination of optical properties of human blood in the spectral range 250 to 1100 nm using Monte Carlo simulations with hematocrit-dependent effective scattering phase functions. J. Biomed. Opt., 11, 034021.

 

Gedzelman, S. D., Lopez-Alvarez, M. A., Hernandez-Andres, J. and Greenler, R. (2000) Quantifying the ‘milky sky” experiment. App. Opt., 47:H128-H132,

 

Ghiradella, H. T. and Butler, M. W. (2009) Many variations on a few themes: a broader look at development of iridescent scales (and feathers). J. R. Soc. Interface, 6:S243-S251

 

Goss, G. A., Hayes, J. A. and Burdon, J. G. W. (1988) Deoxyhemoglobin concentrations in the detection of central cyanosis. Thorax, 43:212-213.

 

Hery, C. (2005) Implementing a Skin BSSRDF (or several ...). SIGGRAPH 2005 Course 9, pp. 14-28.

 

Hidano, A., Ikeda, S., Mizutani (sic), Miyasato, H. and Niimura, M. (1967) Natural history of nevus of Ota. Arch. Derm., 95:187-195.

 

Igarashi, T., Nishino, K. and Nayar, S. K. (2005) The appearance of skin. Technical report, CUCS-024-05, Department of Computer Science, Columbia University, 1-85.

 

Ingram, A. L. and Parker, A. R. (2008) A review of the diversity and evolution of photonic structures in butterflies, incorporating the work of John Huxley (The Natural Museum, London from 1961 to 1990). Phil. Trans. R. Soc. B, 363:2465-2480.

 

Jacques, S. J. (2010) Optical assessment of cutaneous blood volume depends on the vessel size distribution: A computer simulation study. J. Biophton., 3:75-81.

 

Jacques, S. J., Roman, J. R. and Lee K. (2000) Imaging superficial tissues with polarized light. Lasers Surg. Med., 26:119-129.

 

Jung, B., Choi, B., Durkin, A. J. and Kelly, K. M. (2004) Characterization of port wine stain erythema and melanin content using cross-polarized diffuse reflectance imaging. Lasers Surg. Med., 34:174-181.

 

Kerker, M. (1991) Founding fathers of light scattering and surface-enhanced Raman scattering. App. Opt., 30:4699-4705.

 

Kienle, A., Lilge, L., Vitkin, I. A., Patterson, M. S., Wilson, B. C., Hibst, R. and Steiner, R. (1996) Why do veins appear blue? A new look at an old question. App. Opt., 35:1151-1160.

 

Kinoshita, S., Yoshioka, S., Fujii, Y. and Okamoto, N. (2002) Photophysics of structural color in the Morpho butterflies.  Forma, 17:103-121.

 

Kinoshita, S., Yoshioka, S. and Miyazaki, J. (2008) Physics of structural colors.  Rep. Prog. Phys., 71:1-30.

 

Kollias, N., Seo, I. and Bargo, P. R. (2010) Interpreting diffuse reflectance for in vivo skin reactions in terms of chromophores. J. Biophotonics, 3:15-24.

 

Koster, P. H. L., van der Horst, C. M. A. M., Bossyut, P. M. M. and van Gemert, M. J. C. (2001) Lasers Surg. Med., 29:151-155.

 

Krauskoph, J., Zaidi, Q. and Mandler, M. B. (1986) Mechanisms of simultaneous color induction. J. Opt. Soc. Am. A, 3:1752-1757.

 

Lakmaker, O., Pickering, J. W. and Gemert, M. J. C. van (1993) Modeling the perception of port wine stains and its relation to the depth of laser coagulated blood vessels. Lasers Surg. Med., 13:219-226.

 

Lee, E. C. and Park, K. R. (2009) Restoration method of skin scattering blurred vein image for finger vein recognition. Electron. Lett., 45:1074-1076.

 

Lilienfeld, P. (2004) A blue sky history. Optics & Photonics News, June, 32-39.

 

Lister, T., Wright, P. and Chappell, P. (2010) Spectrophotometers for the clinical assessment of port-wine stain skin lesions: a review. Lasers Med. Sci., 25:445-457.

 

Lui, H. and Zhou, Y. (2008) Nevi of Ota and Ito. Site last accessed: 04 May 2010, emedicine.medscape.com/article/1058580-overview.

 

Matsumura, T., Sato-Matsumura, K. C., Yokota, T., Kobayashi, H., Nagashima, K. and Ohkawara, A. (1999) Three-dimensional reconstruction in dermatopathology – A personal computer-based system. J. Cut. Path., 26:197-200.

 

Meinke, M., Muller, G., Helfmann, J. and Freibel, M. (2007) Optical properties of platelets and blood plasma and their influence on the optical behavior of whole blood in the visible to near infrared wavelength range. J. Biomed. Opt., 12, 014024.

 

Meinke, M., Muller, G., Helfmann, J. and Freibel, M. (2007) Empirical model functions to calculate hematocrit-dependent optical properties of human blood. App. Opt., 46:1742-1753.

 

Murali, R., McCarthy, S. W. and Scolyer, R. A. (2009) Blue nevi and related lesions: A review highlighting atypical and newly described variants, distinguishing features and diagnostic pitfalls. Adv. Anat. Pathol.,16:365-382.

 

Mutrynowska, J. and Grzegorzewski (2007) Optical analysis of red blood cell sediment formation. Biorheology, 44:285-297.

 

Nicodemus, F. (1965) Direction reflectance and emissitivity of an opaque surface. App. Opt., 4:767-775.

 

Nishidate, I., Aizu, Y. and Mishina, H. (2004) Estimation of melanin and hemoglobin in skin tissue using multiple regression analysis aided by Monte Carlo simulation. J. Biomed. Opt., 9:700-710.

 

Nishidate, I., Maeda, T., Aizu, Y. and Niizeki, K. (2007) Visualizing depth and thickness of a local blood region in skin tissue using diffuse reflectance images. J. Biomed. Opt., 12, 054006.

 

Poladian, L. Wickham, S., Lee, K. and Large, M. C. J. (2009) Iridescence from photonic Crystals and its suppression in butterfly scales. J. R. Soc. Interface, 6:S233-S242.

 

Prum, R. O. and Torres, R. H. (2004) Structural colouration of mammalian skin: convergent evolution of coherently scattering dermal collagen arrays.  J. Exp. Biol., 207:2157-2172.

 

Rah, D. K., Kim, S. C., Lee, K. H., Park B.Y. and Kim, D. W. (2001) Objective evaluation of treatment effects on port-wine stains using L*a*b* color coordinates. Plast. Reconstr. Surg., 108:842-847.

 

Ramanujam, N. (2000) Fluorescence spectroscopy in vivo.  In: Encyclopedia of Analytic Chemistry, R. A. Meyers (Ed.), John Wiley & Sons, Ltd., Chichester, pp. 20-56.

 

Reisfeld, P. L. (2000) Blue in the skin. A. Acad. Dermatol., 42:597-605.

 

Rodriguez-Diaz, E., Velez-Reyes, M., Chin, R. K. and DiMarzio, C. A. (2000) Wavelength selection for imaging hemoglobin in the skin. SPIE, 4129:243-248.

 

Roggins, A., Friebel, M., Dorschel, K., Hahn, A. and Muller, G. (1999) Optical properties of circulating human blood in the wavelength range 400-2500 nm. J. Biomed. Opt., 4:36-46.

 

Ryan, T. J. (1991) Cutaneous circulation. In: Physiology, Biochemistry, and Molecular Biology of the Skin (2^nd Ed.), L. A. Goldsmith (Ed.), Oxford Press, New York.

 

Saint-Cyr, M., Wong, C., Schaverien, M., Mojallal, A. and Rohrich, R. J. (2009) The perforasome theory: Vascular anatomy and clinical implications. Plast. Reconstr. Surg., 124:1529-1544.

 

Seago, A. E., Brady, P., Vigneron, J.-P. and Schultz, T.D. (2009) Gold bugs and beyond: a review of iridescence and structural colour mechanisms in beetles (Coleoptera).  J. R. Soc. Interface, 6:S165-S184.

 

Shawkey, M. D., Saranathan, V., Palsdottr, H., Crum, J., Ellisman, M. H., Auer, M. and Prum, R. O. (2009) Electron tomography, three-dimensional Fourier analysis and colour prediction of a three-dimensional amorphous biophotonic nanostructure. J. R. Soc. Interface, 6:S213-S220.

 

Smith, G. S. (2008) Summing the molecular contributions to skylight. Am. J. Phys., 76:816-825.

 

Smithies, D. J., van Gemert, M. J. C., Hansen, M. K., Milner, T. E. and Nelson, J. S. (1997) Three-dimensional reconstruction of port wine stain vascular anatomy from serial histological sections. Phys. Med. Biol., 42:1843-1847.

 

Southworth, L. (2007) Ask a geneticist.  Site last accessed: 04 May 2010, www.thetech.org/genetics/ask.php?id=232.

 

Stamatas, G. N. and Kollias, N. (2004) Blood stasis contributions to the perception of skin pigmentation. J. Biomed. Opt., 9:315-322.

 

Steinke, J. M. and Shepherd, A. P. (1988) Diffusion model of the ideal optical absorbance of whole blood. J. Opt. Soc. Am. A, 5:813-822.

 

Swerlick, R. A. (1997) The structure and function of the cutaneous vasculature. J. Dermatol., 24:734-738.

 

Takahashi, S. and Ejima, Y. (1983) Function relationship between chromatic induction and luminance of the inducing stimulus. J. Opt. Soc. Am., 73:198-202.

 

Tang, S. V., Gilchrest, B. A., Noe, J. M., Arndt, K. A., Borgelais, D. B. C. and Itzkan, I. (1983) In vivo spectrophotometric evaluation of normal, lesional, and laser-treated skin in patients with port-wine stains. J. Invest. Dermatol., 80:420-423.

 

Tearney, G. J., Brezinski, M. E., Southern, J.F., Bouma, B.E., Hee, M. R. and Fujimoto, J. G. (1995) Determination of the refractive index of highly scattering human tissue by optical coherence tomography.  Optics Letters, 20:2258-2260.

 

Tsuchida, Y., Fukuda, O. and Kamata, S. (1991) The correlation of skin blood flow with age, total cholesterol, hematocrit, blood pressure, and hemoglobin. Plast. Reconstr. Surg., 88:844-850.

 

van der  Weijer, J., Schmid, C., Verbeek, J. and Larlus, D. (2009) Learning color names for real-world applications. IEEE Trans. Image Proc. 18:1512-1523.

 

Verkruysse, W., Lucassen, G. W., de Boer, J. F., Smithies, D. J., Nelson, J. S. and van Gemert, J. C. (1997) Modeling light distributions of homogeneous versus discrete absorbers in light irradiated turbid media.  Phys. Med. Biol., 42:51-65.

 

Verkruysse, W., Lucassen, G. W. and van Gemert, J. C. (1999) Simulation of color of port wine stain and its dependence on skin variables. Lasers Surg. Med., 25:131-139.

 

Watanabe, T. and Tanaka, T. (2009) Vein authentication using color information and image matching with high performance on natural light. ICROS-SICE International Joint Conference 2009, Fukuoka International Congress Center, Japan.

 

Widdowson, D. C., Moore, J. C., Wright, P. A. and Shakespeare, P. G. (2010) Determination of the effects of blood depth in the dermis on skin colour in a novel skin phantom using digital imaging.  Lasers Med. Sci., 25:55-59.

 

Wikipedia (2010) Eye color. Site last accessed: 04 May 2010, en.wikipedia.org/wiki/Eye_color.

 

Widdowson, D. C., Moore, J. C., Wright, P. A. and Shakespeare, P. G. (2010) Determination of the effects of blood depth in the dermis on skin colour in a novel skin phantom using digital imaging. Lasers Med. Sci., 25:55-59.

 

Xu, F., Lu, T. J., Seffen, K. A. and Ng, E. Y. K. (2009) Mathematical Modeling of Skin Bioheat Transfer. Applied Mechanics Reviews, 62:1-35.

 

Young, A. T. (1982) Rayleigh scattering. Physics Today, January, pp. 42-48.

 

 

Image Credits

 

( 4)      Wikipedia, 2010

( 5)      UC Irvine, 2009

( 9)      Surface Optics, 2009

(10)     Mellado, 1997

(11)     CVI, 2009

(12)     Wikipedia, 2009

(14)     Lim, 2007

Igarashi, 2005

Beckman, 2009

(15)     Melanoma blogsome, 2010

(17)     torynorrhina, living-jewels

butterflies, monzo

hummingbird, Mongabay, David J. Southall, mucubaji.com

structural colors, Kinoshita et al., 2002, 2008, Seago et al., 2009.

(18)     psychedelic fish, about

mandarin fish, ecodiversfeather, Harold Davis, Photonet

blue glacial ice, birdsasart

deep blue sky, wallpapersphere

(19)     collagen, primates, Prum and Torres 2003, 2004

blue eye, grandmother

perioral cyanotic neonate, Stanford

(20)     Blue Nevus, skinmolesblogspot

nevus of Ito, dermatlas

nevus of Ota, medscape

Mongolian Spot, blogspot

(24)     get Kubelka-Munk equations link

(25)     Lihong Wang and Stephen Jacques, 1998

(27)     Braverman et al., 1986

(29)     Smithies et al., 1997

(30)     Matsumura et al., 1999

(34)     www,handprint.com

            www.hugsandscience.com