2018/05/17 by Gabriele Campanella, Campanella, Gabriele, Vitor Werneck Krauss Silva +3 · 5 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · Medicine · #AI in cancer detection #Cell Image Analysis Techniques #Computer Vision and Pattern Recognition (cs.CV) #Digital Imaging for Blood Diseases #FOS: Computer and information sciences #Radiomics and Machine Learning in Medical Imaging
paper · pdf · doi:10.48550/arxiv.1805.06983
openalex publication_date 2018/05/17 · openalex created_date 2022/08/30 · openalex updated_date 2026/07/28
In the field of computational pathology, the use of decision support systems\npowered by state-of-the-art deep learning solutions has been hampered by the\nlack of large labeled datasets. Until recently, studies relied on datasets in\nthe order of few hundreds of slides which are not enough to train a model that\ncan work at scale in the clinic. Here, we have gathered a dataset consisting of\n12,160 slides, two orders of magnitude larger than previous datasets in\npathology and equivalent to 25 times the pixel count of the entire ImageNet\ndataset. Given the size of our dataset it is possible for us to train a deep\nlearning model under the Multiple Instance Learning (MIL) assumption where only\nthe overall slide diagnosis is necessary for training, avoiding all the\nexpensive pixel-wise annotations that are usually part of supervised learning\napproaches. We test our framework on a complex task, that of prostate cancer\ndiagnosis on needle biopsies. We performed a thorough evaluation of the\nperformance of our MIL pipeline under several conditions achieving an AUC of\n0.98 on a held-out test set of 1,824 slides. These results open the way for\ntraining accurate diagnosis prediction models at scale, laying the foundation\nfor decision support system deployment in the clinic.\n