Smartphone camera gives objective assessment of skin tone

19 Jun 2025

Brown University project tackles potential biases in pulse oximetry across racial groups.

The growing use of optical techniques in point-of-care diagnostics has brought with it the need to identify and assess any inherent biases in the results.

One source of these concerns is the differing light absorption properties of different skin types, a factor influencing several promising diagnostic methods.

A 2024 session at SPIE Photonics West on the topic concluded that, as one example, photoacoustic techniques can be "confounded by skin color."

Pulse oximetry, which monitors optical absorption at two wavelengths to determine blood oxygenation, is another technique where the data are complicated by the light absorption of melanin, present in higher levels in darker skin.

A UK report in 2024 urged makers of pulse oximeters to design better, smarter devices that are not biased by skin tone, possibly by developing multi-wavelength systems that are able to correct for the effects of skin pigmentation.

A project at Brown University and Morgan State University has now demonstrated a smartphone-based skin-assessment imaging algorithm that could help standardize skin tone assessment in pulse oximetry.

As discussed in Biophotonics Discovery, the project focused on the determination of individual typology angle (ITA), a colorimetric parameter related to the reflection of light from different skin tones and which can be used to classify skin from very light to dark. The project developed a bespoke algorithm designed to take a designated area of pixels from a 72 dpi smartphone camera image, carry out normalization and gamma correction, and convert device-dependent RGB colors into device-independent ITA data.

Addressing these pernicious issues

"Image-based colorimetric methods are convenient and noninvasive, but accuracy is challenging due to differences in lighting conditions in point-of-care facilities," noted the team in its published paper. "To address these pernicious issues, we present a widely accessible method along with recommendations for skin-tone quantification using a smartphone."

In proof-of-concept trials on four subjects with a diverse distribution of skin pigmentation, images of the index finger were taken from both palmar (palm side) and dorsal (posterior) locations, common spots for pulse oximetry determination. The new algorithm then calculated a standardized ITA measurement, which was compared to those from a professional-grade colorimeter.

Results showed that the smartphone-based readings closely matched those from the high-end instrument. The best results came when both camera flash and room lights were turned off, and the phone was set to a specific exposure level, noted the project. Under these conditions, the method proved consistent across different skin tones and required no extra equipment beyond the phone itself.

The project also formulated specific guidelines for medical practitioners seeking to adopt smartphone-based skin-tone assessments for pulse oximetry research in clinical settings, tackling some concerns identified in the previous studies.

Variations in pigmentation, such as those caused by tattoos, scars, and other skin anomalies, could significantly influence melanin distribution and skin texture or introduce underlying physiological changes, said the group.

Control of ambient lighting and of smartphone settings relating to white balance exposure can minimize automatic adjustments that may skew skin-tone readings, while saving images in a standard format such as jpeg ensures compatibility with analysis software.

"Our findings support the feasibility of using smartphone cameras for skin-tone assessment, specifically in the context of improving pulse oximetry accuracy and addressing skin-tone bias in clinical diagnostics," commented the project. "With additional testing and refinement, this approach has the potential to enhance healthcare equity."