TY - GEN
T1 - Data-dependent nonlinearity analysis in CT denoising CNNs
AU - Wang, Wenying
AU - Tivnan, Matthew
AU - Li, Junyuan
AU - Stayman, Joseph W.
AU - Gang, Jianan
N1 - Funding Information:
This work is supported, in part, by NIH grants R01CA249538 and R01EB027127.
Publisher Copyright:
© 2022 SPIE.
PY - 2022
Y1 - 2022
N2 - Recent years have seen the increasing application of deep learning methods in medical imaging formation, processing, and analysis. These methods take advantage of the flexibility of nonlinear neural network models to encode information and features in ways that can outperform conventional approaches. However, because of the nonlinear nature of this processing, images formed by neural networks have properties that are highly datadependent and difficult to analyze. In particular, the generalizability and robustness of these approaches can be difficult to ascertain. In this work, we analyze a class of neural networks that use only piece-wise linear activation functions. This class of networks can be represented by locally linear systems where the linear properties are highly data-dependent - allowing, for example, estimation of noise in image output via standard propagation methods. We propose a nonlinearity index metric that quantifies the fidelity of a local linear approximation of trained models based on specific input data. We applied this analysis to three example CT denoising CNNs to analytically predict the noise properties in the output images. We found that the proposed nonlinearity metric highly correlates with the accuracy of noise predictions. The analysis proposed in this work provides theoretical understanding of the nonlinear behavior of neural networks and enables performance prediction and quantitation under certain conditions.
AB - Recent years have seen the increasing application of deep learning methods in medical imaging formation, processing, and analysis. These methods take advantage of the flexibility of nonlinear neural network models to encode information and features in ways that can outperform conventional approaches. However, because of the nonlinear nature of this processing, images formed by neural networks have properties that are highly datadependent and difficult to analyze. In particular, the generalizability and robustness of these approaches can be difficult to ascertain. In this work, we analyze a class of neural networks that use only piece-wise linear activation functions. This class of networks can be represented by locally linear systems where the linear properties are highly data-dependent - allowing, for example, estimation of noise in image output via standard propagation methods. We propose a nonlinearity index metric that quantifies the fidelity of a local linear approximation of trained models based on specific input data. We applied this analysis to three example CT denoising CNNs to analytically predict the noise properties in the output images. We found that the proposed nonlinearity metric highly correlates with the accuracy of noise predictions. The analysis proposed in this work provides theoretical understanding of the nonlinear behavior of neural networks and enables performance prediction and quantitation under certain conditions.
UR - http://www.scopus.com/inward/record.url?scp=85131202729&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85131202729&partnerID=8YFLogxK
U2 - 10.1117/12.2612569
DO - 10.1117/12.2612569
M3 - Conference contribution
C2 - 35601024
AN - SCOPUS:85131202729
T3 - Progress in Biomedical Optics and Imaging - Proceedings of SPIE
BT - Medical Imaging 2022
A2 - Zhao, Wei
A2 - Yu, Lifeng
PB - SPIE
T2 - Medical Imaging 2022: Physics of Medical Imaging
Y2 - 21 March 2022 through 27 March 2022
ER -