Multi-Scale Task Multiple Instance Learning for the Classification of Digital Pathology Images with Global Annotations
Niccolò Marini, Sebastian Otálora, Francesco Ciompi, Gianmaria Silvello, Stefano Marchesin, Simona Vatrano, Gianziana Buttafuoco, Manfredo Atzori, Henning Müller · ArODES (HES-SO (https://www.hes-so.ch/)) · 2021
Whole slide images (WSIs) are high-resolution digitized images of tissue samples, stored including di_erent magni_cation levels. WSIs datasets often include only global annotations, available thanks to pathology reports. Global annotations refer to global _ndings in the high-resolution image and do not include information about the location of the regions of interest or the magni_cation levels used to identify a _nding. This fact can limit the training of machine learning models, as WSIs are usually very large and each magni_cation level includes di_erent information about the tissue. This paper presents a Multi-Scale Task Multiple Instance Learning (MuSTMIL) method, allowing to better exploit data paired with global labels and to combine contextual and detailed information identi_ed at several magni_cation levels. The method is based on a multiple instance learning framework and on a multi-task network, that combines features from several magni_cation levels and produces multiple predictions (a global one and one for each magni_cation level involved). MuSTMIL is evaluated on colon cancer images, on binary and multilabel classi_cation. MuSTMIL shows an improvement in performance in comparison to both single scale and another multi-scale multiple instance learning algorithm, demonstrating that MuSTMIL can help to better deal with global labels targeting full and multi-scale images. Keywords: Multi-Scale Multiple Instance Learning, Multiple Instance Learning, Multiscale approach, Computational pathology.