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

M Bengs,S Pant,U Schüller,M Bockmayr,A Schlaefer

doi:10.1515/cdbme-2021-1014

Abstract

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 a microscope to assess the severity of the tumor. This is a timeconsuming task and 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 for classification, while the size of the tiles has not been systematically evaluated. We study the impact of tile size and input strategy and classify the two major histopathological subtypes-Classic and Desmoplastic/Nodular. To this end, we use recently proposed EfficientNets and evaluate tiles with increasing size combined with various downsampling scales. Our results demonstrate using large input tiles pixels followed by intermediate downsampling and patch cropping significantly improves MB classification performance. Our top-performing method achieves the AUC-ROC value of 90.90% compared to 84.53% using the previous approach with smaller input tiles.

Highlights

Medulloblastoma (MB) is the most common malignant brain tumor in children and a major cause of morbidity, as well as mortality in pediatric oncology [15]
Neuropathologists assess the tissue slides under the microscope or digitize the magnified view to obtain an extremely high-resolution image which is called Whole Slide Image (WSI)
We systematically study the effect of tile size, image downsampling, and input strategy for the task of MB subtype classification using pre-trained convolutional neural networks (CNN)

Summary

Introduction

Medulloblastoma (MB) is the most common malignant brain tumor in children and a major cause of morbidity, as well as mortality in pediatric oncology [15]. All MBs are classified as Grade IV tumors by the World Health Organization (WHO) [10], indicating they are invasive and fast-growing. The 2016 edition of the World Health Organization Classification of Tumors of the Central Nervous System (CNS) has defined four histological subtypes of MB [10, 12]— classic type (CMB), desmoplastic/nodular type (DN), MB with extensive nodularity (MBEN), and large cell anaplastic MB (LCA). The visual assessment of such tissue scans is a laborious and timeconsuming task, which is affected by inter-observer variability [2]. These problems have emphasized the requirement of automated decision support tool [17]

Methods

Results

Discussion

Conclusion