Mitigating unknown biases in CT data using machine learning

Junyuan Li; Grace J. Gang; Webster J. Stayman

doi:10.1117/12.2582274

15 February 2021 Mitigating unknown biases in CT data using machine learning

Junyuan Li, Grace J. Gang, Webster J. Stayman

Author Affiliations +

Proceedings Volume 11595, Medical Imaging 2021: Physics of Medical Imaging; 1159541 (2021) https://doi.org/10.1117/12.2582274
Event: SPIE Medical Imaging, 2021, Online Only

Conference Poster

Abstract

CT reconstruction requires an accurate physical model. Mismatches between model and data represent unmodeled biases and can induce artifacts and systematic quantitation errors. Bias effects are dependent on bias structure and data processing (e.g. model-based iterative reconstruction versus filtered-backprojection). In this work, we illustrate this sensitivity and develop a strategy to estimate unmodeled biases for situations where the underlying source is unknown or difficult to estimate directly. We develop a CNN framework for projection-domain de-biasing using a ResUNet architecture and spatial-frequency loss function. We demonstrate a reduction in reconstruction errors across bias conditions and reconstruction methods.

Conference Presentation

Citation Download Citation

Junyuan Li, Grace J. Gang, and Webster J. Stayman "Mitigating unknown biases in CT data using machine learning", Proc. SPIE 11595, Medical Imaging 2021: Physics of Medical Imaging, 1159541 (15 February 2021); https://doi.org/10.1117/12.2582274

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