Study Validates Image-Based, Universal Scoring System for Alopecia Hair Loss

Cameron Gudobba, BA

A recent study found that a newly-developed ‘HairComb’ scoring system showed strong correlations with underlying percentage hair loss, validating the model for future use by alopecia patients.

According to the study investigators, prior to their research there has been no automated system by which hair density can be tracked in order to assess alopecia severity.

The study was led by Cameron Gudobba, BA, from Perelman School of Medicine’s Department of Dermatology at the University of Pennsylvania.

“A tool that could automate hair loss measurement in such a way that correlates closely with existing clinical scales would greatly diminish the time needed to assess the degree of hair loss and would do so in a more consistent and standardized manner,” Gudobba and colleagues wrote. “In this article, we propose that a single quantification system that measures percentage hair loss would be useful across different types of hair loss.”

Research and Methods

The investigators developed HairComb through a new algorithmic quantification system for hair loss in general, using retrospective image data collection to make analyses of alopecia progression in patients. The system finds the hair loss percentage at each pixel, closely automating the Alopecia Density and Extent (ALODEX) score.

The score can also be binarized so as to assess the Severity of Alopecia Tool (SALT) score of the participants’ total scalp. HairComb can assess this after it is given 4 views or the percent- age affected area.

HairComb was described as a convolutional neural network (CNN) designed to combine patients’ overall patterns of hair and pixel-level hair density data starting through the use of a single image. The span of visual characteristics for alopecia covered the range of hair textures, hair colors, and skin tones.

The multicenter research study gathered images from 2015 to 2021 through the University of Pennsylvania and through the Children’s Hospital of Philadelphia, using the interface from Penn Dermatology. The research team recruited 404 study participants from age 2 and older, and scoring systems correlation analysis was used on 250 of these participants.

Study Results

The study results indicated that the newly-automated hair loss percentage resulted in over 92% segmentation accuracy as well as errors that were noted as being comparable to human annotators. For the 250 participants on which a correlation analysis of scoring systems was performed, 70.4% were female and the mean age of participants was 35.3 years.

Of the participants assessed, 75 had androgenetic alopecia (AGA), 66 had alopecia areata (AA), 50 had central centrifugal cicatricial alopecia (CCCA), 27 others had alopecia diagnoses such as frontal fibrosing alopecia and lichen planopilaris, and 32 of them had unaffected scalps with no alopecia.

“While many other scales for alopecias exist, this study focused on analyzing scoring systems for AA, (female-pattern hair loss), and CCCA because they had been clinically validated for grading 2-dimensional photographs that could be also recreated for computational image analysis,” they wrote. “Moving toward 3-dimensional scalp visualizations will enable expanding this work to other alopecia scoring systems.”

The story, “Automating Hair Loss Labels for Universally Scoring Alopecia From Images/ Rethinking Alopecia Scores,” was published online in JAMA Dermatology.