2019/11/13 by Samuel W. Remedios, Zihao Wu, Remedios, Samuel W. +15
Biochemistry, Genetics and Molecular Biology · Computer Science · #Biomedical Text Mining and Ontologies #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Machine Learning in Healthcare
paper · pdf · doi:10.48550/arxiv.1911.05650
openalex publication_date 2019/11/13 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Multiple instance learning (MIL) is a supervised learning methodology that\naims to allow models to learn instance class labels from bag class labels,\nwhere a bag is defined to contain multiple instances. MIL is gaining traction\nfor learning from weak labels but has not been widely applied to 3D medical\nimaging. MIL is well-suited to clinical CT acquisitions since (1) the highly\nanisotropic voxels hinder application of traditional 3D networks and (2)\npatch-based networks have limited ability to learn whole volume labels. In this\nwork, we apply MIL with a deep convolutional neural network to identify whether\nclinical CT head image volumes possess one or more large hemorrhages (>\n20cm3), resulting in a learned 2D model without the need for 2D slice\nannotations. Individual image volumes are considered separate bags, and the\nslices in each volume are instances. Such a framework sets the stage for\nincorporating information obtained in clinical reports to help train a 2D\nsegmentation approach. Within this context, we evaluate the data requirements\nto enable generalization of MIL by varying the amount of training data. Our\nresults show that a training size of at least 400 patient image volumes was\nneeded to achieve accurate per-slice hemorrhage detection. Over a five-fold\ncross-validation, the leading model, which made use of the maximum number of\ntraining volumes, had an average true positive rate of 98.10%, an average true\nnegative rate of 99.36%, and an average precision of 0.9698. The models have\nbeen made available along with source code to enabled continued exploration and\nadaption of MIL in CT neuroimaging.\n