Journal of Investigative Dermatology: May 2021

The presentation of skin disease varies widely across skin types, disease acuity, immune status, and treatment history. Thus, dermatological diagnosis remains challenging.

The broad range of possible skin diseases and presentations of those diseases makes this challenge particularly well suited for applications of artificial intelligence (AI)-based clinical decision support tools.

Using artificial intelligence and machine learning, the application analyzes the lesion type, then provides simple questions to quickly get to a differential diagnosis.

In this study, authors evaluate the performance of a standalone AI tool to correctly categorize a skin lesion's morphology from a test bank of images. To provide a marker of performance, authors evaluate the accuracy of primary care physicians to categorize skin lesion morphology in the same test bank of images without any aids and then with the aid of a simple visual guide.

TAKE HOME MESSAGE

The AI system achieved an accuracy of 68% in determining the single most likely morphology from the test image bank. When the AI’s top prediction was broadened to its top three most likely predictions, accuracy improved to 80%.

In comparison, the diagnostic accuracy of primary care physicians was 36% without any aids and 68% with the visual guide.

The AI was subsequently tested on an additional set of 222 heterogeneous images of varying Fitzpatrick skin types and achieved an overall accuracy of 70% in the Fitzpatrick I–III skin type group and 68% in the Fitzpatrick IV–VI skin type group.

An AI is a powerful tool to assist physicians in the diagnosis of skin lesions while still requiring the user to critically consider other possible diagnoses.

Fundamental to the generation of a dermatological differential is first to classify correctly the primary and secondary morphological features (e.g., patch, plaque, papule, ulcer with or without scale) of any skin disease—a task that can be facilitated by clinical decision support via AI tools.

Overall, the AI tool offered a high level of accuracy when tasked with identifying a broad set of dermatological images ranging from inflammatory to neoplastic conditions and including a broad morphological corpus. 

An AI tool that provides morphological assessment may help users build a differential diagnosis tailored to that individual patient and would be of clinical use in the primary care and emergency settings.