Contrastive self-supervised learning from 100 million medical images with optional supervision.

Autor:	Ghesu FC; Siemens Healthineers, Digital Technology and Innovation, Princeton, New Jersey, United States., Georgescu B; Siemens Healthineers, Digital Technology and Innovation, Princeton, New Jersey, United States., Mansoor A; Siemens Healthineers, Digital Technology and Innovation, Princeton, New Jersey, United States., Yoo Y; Siemens Healthineers, Digital Technology and Innovation, Princeton, New Jersey, United States., Neumann D; Siemens Healthineers, Digital Technology and Innovation, Erlangen, Germany., Patel P; Siemens Healthineers, Digital Technology and Innovation, Princeton, New Jersey, United States., Vishwanath RS; Siemens Healthineers, Digital Technology and Innovation, Bangalore, Karnataka, India., Balter JM; University of Michigan, Department of Radiation Oncology, Ann Arbor, Michigan, United States., Cao Y; University of Michigan, Department of Radiation Oncology, Ann Arbor, Michigan, United States., Grbic S; Siemens Healthineers, Digital Technology and Innovation, Princeton, New Jersey, United States., Comaniciu D; Siemens Healthineers, Digital Technology and Innovation, Princeton, New Jersey, United States.
Jazyk:	angličtina
Zdroj:	Journal of medical imaging (Bellingham, Wash.) [J Med Imaging (Bellingham)] 2022 Nov; Vol. 9 (6), pp. 064503. Date of Electronic Publication: 2022 Nov 30.
DOI:	10.1117/1.JMI.9.6.064503
Abstrakt:	Purpose: Building accurate and robust artificial intelligence systems for medical image assessment requires the creation of large sets of annotated training examples. However, constructing such datasets is very costly due to the complex nature of annotation tasks, which often require expert knowledge (e.g., a radiologist). To counter this limitation, we propose a method to learn from medical images at scale in a self-supervised way. Approach: Our approach, based on contrastive learning and online feature clustering, leverages training datasets of over 100,000,000 medical images of various modalities, including radiography, computed tomography (CT), magnetic resonance (MR) imaging, and ultrasonography (US). We propose to use the learned features to guide model training in supervised and hybrid self-supervised/supervised regime on various downstream tasks. Results: We highlight a number of advantages of this strategy on challenging image assessment problems in radiography, CT, and MR: (1) significant increase in accuracy compared to the state-of-the-art (e.g., area under the curve boost of 3% to 7% for detection of abnormalities from chest radiography scans and hemorrhage detection on brain CT); (2) acceleration of model convergence during training by up to 85% compared with using no pretraining (e.g., 83% when training a model for detection of brain metastases in MR scans); and (3) increase in robustness to various image augmentations, such as intensity variations, rotations or scaling reflective of data variation seen in the field. Conclusions: The proposed approach enables large gains in accuracy and robustness on challenging image assessment problems. The improvement is significant compared with other state-of-the-art approaches trained on medical or vision images (e.g., ImageNet). (© 2022 Society of Photo-Optical Instrumentation Engineers (SPIE).)
Databáze:	MEDLINE
Externí odkaz:	Zobrazit plný text záznamu