CNN-Based Projected Gradient Descent for Consistent CT Image Reconstruction

By Harshit Gupta, Kyong Hwan Jin, Ha Quy Nguyen, Michael T. McCann, Michael A. Unser. IEEE Transactions on Medical Imaging, 2017

397 citations recorded by the source corpus.

We present a new image reconstruction method that replaces the projector in a projected gradient descent (PGD) with a convolutional neural network (CNN). Recently, CNNs trained as image-to-image regressors have been successfully used to solve inverse problems in imaging. However, unlike existing iterative image reconstruction algorithms, these CNN-based approaches usually lack a feedback mechanism to enforce that the reconstructed image is consistent with the measurements. We propose a relaxed version of PGD wherein gradient descent enforces measurement consistency, while a CNN recursively projects the solution closer to the space of desired reconstruction images. We show that this algorithm is guaranteed to converge and, under certain conditions, converges to a local minimum of a non-convex inverse problem. Finally, we propose a simple scheme to train the CNN to act like a projector. Our experiments on sparse-view computed-tomography reconstruction show an improvement over total variation-based regularization, dictionary learning, and a state-of-the-art deep learning-based direct reconstruction technique.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/aba7570d73f8a6d320ec7830e69d50ac1899c17c

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

GAN-Based Projector for Faster Recovery With Convergence Guarantees in Linear Inverse Problems (2019, background): “However, [1, 12, 15, 25] use information about the measurement matrix A while training the network.”
Algorithm Unrolling: Interpretable, Efficient Deep Learning for Signal and Image Processing (2019, result): “In a similar fashion, in [67] Gupta et al.”
ADMM-CSNet: A Deep Learning Approach for Image Compressive Sensing (2020, methodology): “The methods in [61], [62], [63] unroll a primal-dual optimization method or a (projected) gradient descent method for low-dose”
Small Sample Learning in Big Data Era (2018, methodology): “, 2017), sparse view/data Computed Tomography(CT) reconstruction (Gupta et al., 2018; Chen et al., 2017; Adler and Öktem, 2018), compressive image reconstruction (Metzler et al.”
Learning to Optimize: A Primer and A Benchmark (2021, background): “Th works [175; 153; 76; 256]use data-driven training to learn a proximal or a projection operator and apply it in the unrolled algorithm.”
Deep Learning Techniques for Inverse Problems in Imaging (2020, methodology): “Improved methods for training deep neural networks for PnP can be found in [81], [106], [118].”
Deep Convolutional Framelet Denosing for Low-Dose CT via Wavelet Residual Network (2017, background): “Compared to the recent proposals of learning-based optimization approaches [16], [24], one of the important advantages of our work is that our deep network is no more a black box but can be optimized for speciﬁc restoration tasks by choosing optimal framelet representation.”