Supplementary files for "Deep learning based diagnostic triage classifier for melanocytic skin pathology utilizing customizable automated attention to biologically relevant regions"
Description
Supplementary files to accompany the JAAD manuscript "Deep learning based diagnostic triage classifier for melanocytic skin pathology utilizing customizable automated attention to biologically relevant regions". 1. Mendeley Supplementary Methods. Slide digitalization and pre-processing steps, cohort selection and details, cluster development, model design, model implementation and ethics board approval details. 2. Mendeley Supplementary Results. Results from the cluster development stage, and model performance across cohorts and tumor subtypes. 3. Mendeley Supplementary Data 1. Powerpoint file showing the 169 tiles from each cluster that were used to derive a morphological correlate for that cluster. 4. Mendeley Supplementary Data 2A-2F. Excel file with details of the cluster development cohort tumor subtypes (2A), and showing the comparative results between the tumor-associated clusters (3, 6, 17, 18, 19 and 20) and the manual annotations from melanomas (n=85 WSIs) and nevi (n=80 WSIs) in the cluster development cohort (2B-2C), and the clinicopathological features and performance results from the MIA, Leeds and Mexico cohorts using the tumor-associated clusters model (2D-2F).