The zeal to learn about artificial intelligence (AI) and machine learning (ML) has grown in recent years. The dermatopathological field has also witnessed its impact, like the attempt to develop and orient algorithms that can assist the pathologist in the differential diagnosis of complex melanocytic lesions. 

AI and ML are also attractive due to their ability to learn from their mistakes, self-correct and eventually provide valid support to dermatopathologists. 

Although these tools cannot completely replace pathologists, they can greatly assist in histopathological diagnosis. Contrary, the dermatopathological diagnosis encompasses a complete file of clinical information, like the patient′s age, lesion′s topography, dermoscopic features, color variations and ulceration, which the neural network does not document as it is based fundamentally on image discrimination. 

Further, it is also crucial to recognize that dermatopathology is a landscape of diagnostic debate among the pathologists and existence of atypical lesions which can’t be easily framed in defined criteria (like certain types of Spitz nevi, atypical nevi/Spitz tumors, a melanocytic proliferation of sceptical clinical importance like SAMPUS and MELTUMP and variants of melanoma such as Spitzoid melanoma), thus still can′t be concluded in “real life”, further causing difficulty in training a convolutional neural network in the certain distinction of an entity. 

Lastly, it is crucial to acknowledge that any AI algorithm is to be trained by a “human being” and, thus, will always need the “involvement” of the pathologist.

New studies and discoveries in the underexplored areas like dermatopathological diagnosis served by AI are the need of the time.

Source: Cazzato G, Colagrande A, Cimmino A, et al. Artificial intelligence in dermatopathology: new insights and perspectives. Dermatopathology (Basel). 2021;8(3):418-25.