Note: RPI put out a press release today (4/29/2010) on my work today so I am reposting this with today's date also.

If you ever have a mysterious skin rash and show your doctor, he or she may very well take a marker and encircle it, and then ask you to come back in a week. What’s the marker for? That’s obvious. If the rash is expanding, it may grow by only, say, 5% in a week, far too little to notice. But with the marker pen tightly encircling the original extent of the rash, any growth in the rash will become perceptually obvious because it will have grown beyond the boundary of the marker pen. 

As obvious as this is for rashes, this basic principle is not currently followed for the much more acutely important skin color changes that occur in various medical conditions. Despite skin pallor having had a long history within medicine diagnosis, the medical community is currently handicapped in its ability to visually sense clinically relevant skin color modulations. For skin color, medicine is markerless. If a doctor sees a patient, and then sees the patient again later, the doctor will have little or no idea whether the skin has shifted. Quantitatively small color shifts can have tremendous medical implications: it can, for example, mean the difference between having healthy oxygen saturation and life-threateningly low oxygen saturation. 

But we can do better. Our eyes have evolved, so I have argued in my research, to be near optimally sensitive to skin color changes due to underlying physiological changes in the blood. The skin color changes that mattered over evolution were socio-sexual signals, and socio-sexual signals tend to have strong spectral gradients our oximeter eyeballs are great at detecting. In fact, our eyes depend on these gradients to see skin colors. For example, our perception of blue-green veins depends crucially on its contrast with the surrounding skin. Blue-green veins if viewed all by themselves (i.e., through an aperture) do not appear blue-green at all. Veins are just slightly spectrally shifted toward blue-green compared to the skin’s baseline color, but when seen with the baseline color in the spatial surround, one sees the veins as genuinely blue-green. 

Clinical color changes, on the other hand, are not selected to be seen – they are just side effects of being sick – and can often lead to much more spatially uniform shifts in color. And that’s the problem. If a patient’s skin color shifts a small amount for many clinical reasons (like central cyanosis), then when the doctor or nurse comes back, the color shift will often be imperceptible.
We can, however, fix this with a little marker pen. Rather than encircling the baseline extent of the rash, we need a marker pen to record the baseline color of the patient’s skin. And one place to begin marking up is the beloved hospital gown (although the same point will apply to any other colors visually proximal to the patient, such as the walls or sheets). The problem with hospital gowns is not just the drafty bum, but that gown colors are not designed to harness the oximetric, blood-diagnosing powers of our eyes.

The figure below shows four sample hospital gown colors, and on each of these colors I have placed the same five patches of skin. The central patch, let us presume, is your baseline skin color. The patches around the central patch are slightly color-shifted from the baseline, namely (from top, clockwise) bluer, redder, yellower and greener. Even with the baseline pitch there in the image, the color shifts are not perceptually salient. All the patches look qualitatively like the baseline patch’s color. And in real life the nurse or doctor would see the central patch, and then come back later and see just one of the shifted colors – determining if the color has changed would be nearly impossible. 

mark changizi nude colored gown color changes

But now consider what happens if I lay those same five colored skin patches adjacent to a gown that has a color matching the baseline skin color in the center. In the figure below the four color-shifted patches look utterly and qualitatively distinct. They don’t look just a wee bit bluer, redder, yellower and greener than baseline. Rather, they now look distinctly blue, red, yellow and green. Again, these are the exact same four patches as were in each of the “gowns” in the previous figure. 

mark changizi nude colored gown color changes

Here, then, is the simple prescriptive advice for hospitals: buy skin tone colored gowns, with colors relevant for the population you serve, and have patients wear a gown that best matches their baseline skin color. The tighter the color match, the more hypersensitive the eyes of the nurses and doctors are at sensing tiny spectral shifts away from baseline. I also suggest “skin color adhesive tabs” in a recent publication, which you can read here:

Are these ideas relevant, you might ask, given that hospitals these days have cheap access to oximeters? Here is what co-author Kevin Rio and I write about this at the end of the article I just mentioned above:
One may wonder if these two techniques for harnessing color vision for oximetry are relevant in modern medicine, given the availability of pulse oximetry. [T]he clinical disciplines most utilizing pulse oximetry are also the disciplines that most often refer to the patient’s clinical skin color in diagnosis: thus, the actual practice of medicine appears to value our human color capabilities, despite the presence of pulse oximetry. There are several potential explanations for this: (a) color perception provides redundant detection of oxygen saturation (e.g., if the oximeter becomes unattached), (b) observation of skin color modulations may lead to a faster behavioral response by the clinician (the “look” of sickness may be more psychologically engaging than numbers or beeps from an oximeter), and (c) our color perception is capable of sensing the spatial gradients in skin color across the body, and the nature of those gradients can impart information to a clinician. Furthermore, there are circumstances where pulse oximetry is not used today, but where the “color oximetry” techniques above would be of great value: (i) in certain parts of the hospital (e.g., in transit, or the emergency department waiting room), (ii) in third world hospitals, where it is still not part of standard care, (iii) in the field (e.g., for athletes or soldiers, and (iv) in the home (e.g., for SIDS detection). 
Haiti comes to mind.