JMIR Publications

ABSTRACT

The performance of deep learning image recognition models is below par when applied to images with Fitzpatrick classification of skin type 4 and 5. The objective of this research was to assess whether image recognition models perform differently when differentiating between dermatological diseases in individuals of skin of color (Fitzpatrick skin type 4 and 5) than when differentiating between the same dermatological diseases in Caucasians (Fitzpatrick skin type 1,2 and 3) when both models are trained on the same number of images. The image recognition models were trained and validated on images using TensorFlow as a deep learning framework and a deep convolutional neural network. The image recognition models trained and validated on images with light skin color had higher sensitivity, specificity, positive predictive value, negative predictive value and F1 score than the image recognition models trained and validated on images of skin of color for differentiation between BCC and melanoma. A higher number of images of dermatological diseases in individuals with darker skin color than images of dermatological diseases in individuals with light skin color would need to be gathered for the AI models to perform equally well.