Deep BCD-Net Using Identical Encoding-Decoding CNN Structures for Iterative Image Recovery

Abstract

In "extreme" computational imaging that collects extremely undersampled or noisy measurements, obtaining an accurate image within a reasonable computing time is challenging. Incorporating image mapping convolutional neural networks (CNN) into iterative image recovery has great potential to resolve this issue. This paper 1) incorporates image mapping CNN using identical convolutional kernels in both encoders and decoders into a block coordinate descent (BCD) signal recovery method and 2) applies alternating direction method of multipliers to train the aforementioned image mapping CNN. We refer to the proposed recurrent network as BCD-Net using identical encodingdecoding CNN structures. Numerical experiments show that, for a) denoising low signal-to-noise-ratio images and b) extremely undersampled magnetic resonance imaging, the proposed BCD-Net achieves significantly more accurate image recovery, compared to BCD-Net using distinct encoding-decoding structures and/or the conventional image recovery model using both wavelets and total variation.

Transform Learning for Magnetic Resonance Image Reconstruction: From Model-Based Learning to Building Neural Networks

“…Some of these works incorporate the measurement forward model (physics) in the reconstruction model that is typically an unrolled iterative algorithm [8][9][10]. Supervised learning of TL-MRI models has also shown promise [9,33].…”

“…Block Supervised Low-Rank Fixed Directional DL TL Matching Learning Modeling Sparse MRI [5] PBDW [19] LORAKS [23] PANO [20] DLMRI [6] SOUPDIL-MRI [28] LASSI [22] STL-MRI [24] FRIST-MRI [27] STROLLR-MRI [12] ADMM-Net [8] BCD-Net [9,33]…”

Joint low-count PET/CT segmentation and reconstruction with paired variational neural networks

“…To further improve denoising capability, we extend the convolutional autoencoder in (5) [23], by replacing {c k } with separate decoding filters {d k }. We define the following updates for each layer:…”

BCD-Net for Low-Dose CT Reconstruction: Acceleration, Convergence, and Generalization

“…There exist several ways to prevent NNs from overfitting, e.g., increasing the dataset size, reducing the neural network complexity, and dropout. However, in This section modifies the architecture of BCD-Net in [4] for CT reconstruction. For the image denoising modules, we use layer-wise autoencoding CNNs that apply exponential function to trainable thresholding parameters.…”

“…Many existing works can be viewed as a special case of BCD-Net. For example, RED [11] and MoDL [1] are special cases of BCD-Net, because they use identical image denoising modules across layers or only consider quadratic data-fidelity terms (e.g., the first term in (P1)) in their MBIR modules.This paper modifies BCD-Net that uses convolutional autoencoders in its denoising modules [4], and applies the modified BCD-Net to low-dose CT reconstruction. First, for fast CT reconstruction, we apply the Accelerated Proximal Gradient method using a Majorizer (APG-M), e.g., FISTA [2], to MBIR modules using the statistical CT data-fidelity term.…”