Estimation of attenuator mask from region of interest (ROI) dose-reduced images for brightness equalization using convolutional neural networks

S. V. Setlur Nagesh; D. R. Bednarek; S. Rudin

doi:10.1117/12.2512646

1 March 2019 Estimation of attenuator mask from region of interest (ROI) dose-reduced images for brightness equalization using convolutional neural networks

S. V. Setlur Nagesh, D. R. Bednarek, S. Rudin

Author Affiliations +

Proceedings Volume 10948, Medical Imaging 2019: Physics of Medical Imaging; 1094845 (2019) https://doi.org/10.1117/12.2512646
Event: SPIE Medical Imaging, 2019, San Diego, California, United States

Abstract

Region-of-Interest (ROI) fluoroscopy uses a differential x-ray attenuator to reduce the dose to the patient in the periphery region while maintaining regular dose within the ROI treatment area, resulting in an image with differential brightness regions. The brightness difference can be corrected by subtracting a mask of the ROI attenuator from the dose-reduced image. The purpose of this work is to implement a Convolutional-Neural-Network (CNN) capable of deriving a mask of the ROI attenuator from the dose-reduced images, which can be used to equalize the brightness in the dose-reduced images without a pre-acquired mask. A data set of 10,000 ROI dose reduced images of various objects including anthropomorphic head and chest phantoms were generated with different ROI positions and sizes. A 22 layer CNN was developed to derive a mask of the ROI attenuator from the dose reduced image. The network was trained on 30% and tested on 15% of the images from the ROI image data set. The trained CNN was used on the remaining 55% of the data set to generate the ROI mask, and the average computation time for each image was calculated to be 70 ms. The Mean-Square-Error (MSE) for the testing data set was calculated to be 2.03e-05. For the remaining data set of 5500 images the average MSE between the network output and the corresponding expected ROI mask was calculated to be 2.21e-05. The masks generated by the CNN were successfully used to restore and equalize the brightness of the dose reduced images.

Citation Download Citation

S. V. Setlur Nagesh, D. R. Bednarek, and S. Rudin "Estimation of attenuator mask from region of interest (ROI) dose-reduced images for brightness equalization using convolutional neural networks", Proc. SPIE 10948, Medical Imaging 2019: Physics of Medical Imaging, 1094845 (1 March 2019); https://doi.org/10.1117/12.2512646

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