Deep Iteration Assisted by Multi-level Obey-pixel Network Discriminator (DIAMOND) for Medical Image Recovery

TL;DR

DIAMOND is a unified deep learning framework combining GANs and iterative optimization to improve medical image restoration, effectively recovering structures and details while suppressing artifacts.

Contribution

It introduces a novel strategy integrating GANs with multi-level obey-pixel networks and ADMM optimization for enhanced medical image recovery.

Findings

Achieves superior image quality in medical restoration tasks.

Effectively suppresses artifacts and recovers subtle details.

Demonstrates robustness across various medical imaging scenarios.

Abstract

Image restoration is a typical ill-posed problem, and it contains various tasks. In the medical imaging field, an ill-posed image interrupts diagnosis and even following image processing. Both traditional iterative and up-to-date deep networks have attracted much attention and obtained a significant improvement in reconstructing satisfying images. This study combines their advantages into one unified mathematical model and proposes a general image restoration strategy to deal with such problems. This strategy consists of two modules. First, a novel generative adversarial net(GAN) with WGAN-GP training is built to recover image structures and subtle details. Then, a deep iteration module promotes image quality with a combination of pre-trained deep networks and compressed sensing algorithms by ADMM optimization. (D)eep (I)teration module suppresses image artifacts and further recovers subtle image details, (A)ssisted by (M)ulti-level (O)bey-pixel feature extraction networks (D)iscriminator to recover general structures. Therefore, the proposed strategy is named DIAMOND.