JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) has emerged in dermatology with some studies focusing on skin disorders, such as skin cancer, atopic dermatitis, psoriasis and onychomycosis. Alopecia areata (AA) is a dermatological disease, which prevalence is 0.7-3% in the US, characterized by oval areas of non-scarring hair loss of the scalp or body without evident clinical variables to predict its response to the treatment. Nonetheless, some studies suggest a predictive value of trichoscopic features in the evaluation of treatment response. Assuming that black dots, broken hairs, exclamation mark and tapered hairs are markers of negative predictive value to the treatment response while yellow dots are markers of no response to the treatment according to recent studies, then the absence of these trichoscopic features could indicate a favourable disease evolution without treatment or even predict its response. Nonetheless, no studies have been reported evaluating the role of AI in AA by using trichoscopic features.

Objective:

To develop an AI algorithm able to predict, by using trichoscopic images, those patients diagnosed with AA with a better disease evolution.

Methods:

80 trichoscopic images were included and classified in those with or without features of negative prognosis. Using a data augmentation technique, they were multiplied to 179 images to training an artificial intelligence algorithm, as has been done with dermoscopic images of skin tumors with good response . Subsequently, 82 new imagenes of alopecia areata were presented to the algorithm and were who classified them as responders and non-responders, this process was reviewed by an expert trichologist observer presenting a concordance higher to 90% with the algorithm identifying structures described previously. The evolution of the cases was followed to truly stablished their response to the treatment and therefore to assess the predictive value of the algorithm.

Results:

32/40 images (80%) predicted as non-responders scarcely showed response to the treatment while 34/42 images (81%) predicted as responders showed good response to the treatment.

Conclusions:

To develope an AI algorithm tool could be useful to predict the AA evolution and its response to the treatment. However, further research is needed including bigger sample images or trained algorithms by using images previously classified according to the disease evolution and not to the trichoscopic features.