Multi-Scale Input Strategies for Medulloblastoma Tumor Classification using Deep Transfer Learning

Abstract Medulloblastoma (MB) is a primary central nervous system tumor and the most common malignant brain cancer among children. Neuropathologists perform microscopic inspection of histopathological tissue slides under microscope to assess severity tumor. This timeconsuming task often infused with observer variability. Recently, pre-trained convolutional neural networks (CNN) have shown promising results for MB subtype classification. Typically, high-resolution images are divided into smaller tiles classification, while size has not been systematically evaluated. We study impact tile input strategy classify two major subtypes-Classic Desmoplastic/Nodular. To this end, we use recently proposed EfficientNets evaluate increasing combined various downsampling scales. Our demonstrate using large pixels followed by intermediate patch cropping significantly improves classification performance. top-performing method achieves AUC-ROC value 90.90% compared 84.53% previous approach tiles.