Track

Clinical Studies

Abstract

Kinase fusions are increasingly identified in Spitz nevi and can influence the clinical behavior and histopathologic features of these melanocytic neoplasms. Unique kinase fusions continue to be discovered which can make clinicopathologic correlation with specific molecular aberrations challenging. Artificial intelligence has recently been explored in dermatopathology to assist with diagnostic evaluation, but the ability of this technology to accurately describe characteristic features of melanocytic tumors with kinase fusions is not well known. Here we present four patients with atypical Spitz melanocytic neoplasms with kinase fusions, including ALK (n=1) and TRK (n=3) fusion mutations. We compare the clinicopathologic features of these tumors and assess a popular artificial intelligence, ChatGPT4 by Open AI, to correctly describe these features and to make educational diagrams of these histopathologic characteristics. ChatGPT4 correctly identified distinctive histopathologic features associated with these kinase fusions, including the plexiform architecture commonly associated with ALK fusions as well as the filigree-like architecture associated with TRK fusions. However, the ChatGPT4 Image Generator was unable to create representative histopathologic models of these characteristic features despite detailed prompts and feedback, instead creating inaccurate cartoonish diagrams. Overall, our study demonstrates the utility of ChatGPT4 in accurately describing key characteristics of atypical Spitz tumors with various kinase fusions and demonstrates this technology may assist in diagnosing these tumors, though its image generation feature is currently unable to create realistic diagrams for educational purposes.