Empirical neural network forward model for maximum likelihood material decomposition in spectral CT

Kevin C. Zimmerman; Adam Petschke

doi:10.1117/12.2255953

9 March 2017 Empirical neural network forward model for maximum likelihood material decomposition in spectral CT

Kevin C. Zimmerman, Adam Petschke

Proceedings Volume 10132, Medical Imaging 2017: Physics of Medical Imaging; 101323S (2017) https://doi.org/10.1117/12.2255953
Event: SPIE Medical Imaging, 2017, Orlando, Florida, United States

Abstract

CT measurements using photon counting detectors provide spectral information that can be used to estimate a material's composition. This material decomposition task is complicated by pulse pileup and charge-sharing phenomena. Physics-based methods that use maximum likelihood to estimate a material's composition rely on accurate modeling of the forward spectral measurement process, including the source spectrum and detector response. An empirical projection-domain decomposition method is proposed that uses energy-bin measurements from known basis material path lengths. The known basis material path lengths and energy-bin measurements are used to train a neural network to model the forward spectral measurement process. The neural network is used with a maximum likelihood algorithm to estimate basis material path lengths with optimal noise properties. The method does not require a model of the source spectrum or detector response. Simulations of a step-wedge phantom containing 10 path lengths of polymethyl methacrylate and 10 path lengths of aluminum resulted in 100 calibration measurements for training. Path lengths not included in calibration were used to evaluate the estimator's performance. Projections of the test path lengths contained 1000 Poisson noise realizations and the bias and variance of the estimated path lengths were used as evaluation metrics. The proposed method had less than 2% bias in the test path lengths and had a variance that achieved the Cramèr-Rao lower bound. The proposed method is an efficient estimator that estimates basis material path lengths with optimal noise properties.

Citation Download Citation

Kevin C. Zimmerman and Adam Petschke "Empirical neural network forward model for maximum likelihood material decomposition in spectral CT", Proc. SPIE 10132, Medical Imaging 2017: Physics of Medical Imaging, 101323S (9 March 2017); https://doi.org/10.1117/12.2255953

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

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
6 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Neural networks

Calibration

Sensors

Aluminum

Data modeling

Photon counting

Signal attenuation

Show All Keywords

Keywords/Phrases

Search In:

Publication Years