Design, training, and applications of foundation model for chest computed tomography volumes

Amara Tariq; Bhavik N. Patel; Imon Banerjee

doi:10.1117/12.3003042

2 April 2024 Design, training, and applications of foundation model for chest computed tomography volumes

Amara Tariq, Bhavik N. Patel, Imon Banerjee

Proceedings Volume 12926, Medical Imaging 2024: Image Processing; 1292611 (2024) https://doi.org/10.1117/12.3003042
Event: SPIE Medical Imaging, 2024, San Diego, California, United States

Abstract

Self-supervised pretraining can reduce the amount of labeled training data needed by pre-learning fundamental visual characteristics of the imaging data. We developed a foundation model for chest computed tomography exams using selfsupervised training strategy of masked image region prediction on 1M chest CT slices. The model was evaluated for two downstream tasks; pulmonary embolism (PE) detection (classification) and lung nodule segmentation. Use of the foundation model as a backbone improved performance and reduced computational effort needed for downstream tasks compared to task-specific state-of-the-art (SOTA) models. PE detection was improved for training dataset sizes as large as 380K with maximum gain of 5% over SOTA. Segmentation model initialized with foundation model weights learned twice as fast as randomly initialized model. This model can then be finetuned with limited task-specific annotated data for a variety of downstream imaging tasks thus accelerating research in biomedical imaging informatics.

Conference Presentation

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Amara Tariq, Bhavik N. Patel, and Imon Banerjee "Design, training, and applications of foundation model for chest computed tomography volumes", Proc. SPIE 12926, Medical Imaging 2024: Image Processing, 1292611 (2 April 2024); https://doi.org/10.1117/12.3003042

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available