Researchers at the University of Leeds developed an artificial intelligence (AI) model capable of identifying invasive melanoma in whole slide images (WSIs) and extracting five objective morphological parameters that independently predict patient survival. The study, published in *The Journal of Pathology: Clinical Research*, highlights the potential of AI to enhance prognostic assessment beyond the current staging system, which relies heavily on the subjective Breslow thickness measurement—a metric unchanged for 56 years. The AI model, a convolutional neural network (CNN), was trained to segment invasive melanoma and subsequently analyzed tumor morphology, deriving parameters such as area, perimeter, major and minor axis lengths, and a novel Nodularity Index (NI), which assesses tumor shape by comparing depth to width. Analysis of 745 WSIs revealed that all five parameters independently predicted overall and melanoma-specific survival, with the minor axis length—renamed "digital Breslow thickness"—showing the strongest association (HR 2.04). The NI, reflecting tumor roundness, was particularly significant, improving prognostic accuracy when combined with existing staging (AUC increase from 0.73 to 0.75). These findings suggest that AI can uncover previously unrecognized morphological biomarkers, addressing the 26% of melanoma survival variance unexplained by current staging. While retrospective design and exclusion of rare subtypes limit generalizability, the study demonstrates AI’s potential to automate prognostic analysis without increasing pathologist workload. Future work includes validation in independent cohorts and prospective trials. The research, supported by UK funding bodies, underscores AI’s role in advancing precision oncology by extracting novel prognostic insights from histopathological data.