A Weakly Supervised Deep Learning Framework for COVID-19 CT Detection and Analysis

Autor:	Maayan Frid-Adar, Rula Amer, Dor Amran, Asher Kabakovitch, Nimrod Sagie, Hayit Greenspan, Ophir Gozes
Rok vydání:	2020
Předmět:	Coronavirus disease 2019 (COVID-19) Computer science business.industry Feature vector Deep learning Chest ct Machine learning computer.software_genre Pipeline (software) Lung segmentation Digital image processing Artificial intelligence Cluster analysis business computer
Zdroj:	Thoracic Image Analysis Lecture Notes in Computer Science Thoracic Image Analysis ISBN: 9783030624682 TIA@MICCAI
ISSN:	1611-3349 0302-9743
DOI:	10.1007/978-3-030-62469-9_8
Popis:	The outbreak of the COVID-19 global pandemic has affected millions and has a severe impact on our daily lives. To support radiologists in this overwhelming challenge, we developed a weakly supervised deep learning framework that can detect, localize, and quantify the severity of COVID-19 disease from chest CT scans using limited annotations. The framework is designed to rapidly provide a solution during the initial outbreak of a pandemic when datasets availability is limited. It is comprised of a pipeline of image processing algorithms which includes lung segmentation, 2D slice classification, and fine-grained localization. In addition, we present the Coronascore bio-marker which corresponds to the severity grade of the disease. Finally, we present an unsupervised feature space clustering which can assist in understanding the COVID-19 radiographic manifestations. We present our results on an external dataset comprised of 199 patients from Zhejiang province, China.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_dedup___::e2f79e356a23804003cb6cbdbaed7dfc Zobrazit plný text záznamu