Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DiffusionXRay: A Diffusion and GAN-Based Approach for Enhancing Digitally Reconstructed Chest Radiographs

Aryan Goyal, Ashish Mittal, Pranav Rao, Manoj Tadepalli, Preetham Putha

TL;DR

DiffusionXRay is introduced, a novel image restoration pipeline for Chest X-ray images that synergistically leverages denoising diffusion probabilistic models (DDPMs) and generative adversarial networks (GANs) that demonstrates promising results in enhancing image clarity, contrast, and overall diagnostic value of chest X-rays while preserving subtle yet clinically significant artifacts.

Abstract

Deep learning-based automated diagnosis of lung cancer has emerged as a crucial advancement that enables healthcare professionals to detect and initiate treatment earlier. However, these models require extensive training datasets with diverse case-specific properties. High-quality annotated data is particularly challenging to obtain, especially for cases with subtle pulmonary nodules that are difficult to detect even for experienced radiologists. This scarcity of well-labeled datasets can limit model performance and generalization across different patient populations. Digitally reconstructed radiographs (DRR) using CT-Scan to generate synthetic frontal chest X-rays with artificially inserted lung nodules offers one potential solution. However, this approach suffers from significant image quality degradation, particularly in the form of blurred anatomical features and loss of fine lung field structures. To overcome this, we introduce DiffusionXRay, a novel image restoration pipeline for Chest X-ray images that synergistically leverages denoising diffusion probabilistic models (DDPMs) and generative adversarial networks (GANs). DiffusionXRay incorporates a unique two-stage training process: First, we investigate two independent approaches, DDPM-LQ and GAN-based MUNIT-LQ, to generate low-quality CXRs, addressing the challenge of training data scarcity, posing this as a style transfer problem. Subsequently, we train a DDPM-based model on paired low-quality and high-quality images, enabling it to learn the nuances of X-ray image restoration. Our method demonstrates promising results in enhancing image clarity, contrast, and overall diagnostic value of chest X-rays while preserving subtle yet clinically significant artifacts, validated by both quantitative metrics and expert radiological assessment.

DiffusionXRay: A Diffusion and GAN-Based Approach for Enhancing Digitally Reconstructed Chest Radiographs

TL;DR

DiffusionXRay is introduced, a novel image restoration pipeline for Chest X-ray images that synergistically leverages denoising diffusion probabilistic models (DDPMs) and generative adversarial networks (GANs) that demonstrates promising results in enhancing image clarity, contrast, and overall diagnostic value of chest X-rays while preserving subtle yet clinically significant artifacts.

Abstract

Deep learning-based automated diagnosis of lung cancer has emerged as a crucial advancement that enables healthcare professionals to detect and initiate treatment earlier. However, these models require extensive training datasets with diverse case-specific properties. High-quality annotated data is particularly challenging to obtain, especially for cases with subtle pulmonary nodules that are difficult to detect even for experienced radiologists. This scarcity of well-labeled datasets can limit model performance and generalization across different patient populations. Digitally reconstructed radiographs (DRR) using CT-Scan to generate synthetic frontal chest X-rays with artificially inserted lung nodules offers one potential solution. However, this approach suffers from significant image quality degradation, particularly in the form of blurred anatomical features and loss of fine lung field structures. To overcome this, we introduce DiffusionXRay, a novel image restoration pipeline for Chest X-ray images that synergistically leverages denoising diffusion probabilistic models (DDPMs) and generative adversarial networks (GANs). DiffusionXRay incorporates a unique two-stage training process: First, we investigate two independent approaches, DDPM-LQ and GAN-based MUNIT-LQ, to generate low-quality CXRs, addressing the challenge of training data scarcity, posing this as a style transfer problem. Subsequently, we train a DDPM-based model on paired low-quality and high-quality images, enabling it to learn the nuances of X-ray image restoration. Our method demonstrates promising results in enhancing image clarity, contrast, and overall diagnostic value of chest X-rays while preserving subtle yet clinically significant artifacts, validated by both quantitative metrics and expert radiological assessment.
Paper Structure (13 sections, 7 equations, 3 figures, 2 tables)

This paper contains 13 sections, 7 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Datasets
  4. Synthesizing Chest X-Ray via Projection from CT Scans
  5. MUNIT-LQ
  6. Loss Functions
  7. DDPM
  8. Experiments
  9. Setup
  10. Results
  11. Automated metrics
  12. Human Evaluation
  13. Conclusion

Figures (3)

  • Figure 1: Overview of our proposed DiffusionXRay framework. (a) Unpaired high and low-quality (CXRs) serve as training data for our unsupervised domain transfer models. (b) Implementation of domain transfer using: (b-1) MUNIT-LQ for style-guided translation or(b-2) DDPM-LQ for diffusion-based degradation modeling using 2 stage training. (c) The paired data generated from these models is subsequently used to train our final enhancement network which is DDPM-HQ.
  • Figure 2: Qualitative comparison of image enhancement and super-resolution results (512$\times$512 $\rightarrow$ 1024$\times$1024) on synthetically degraded chest X-ray (CXR) images. Each input image was a high-quality (HQ) CXR artificially degraded using the MUNIT-LQ framework. The bicubic baseline refers to a DDPM model trained on paired data generated with bicubic downsampling. The reference shown is the original HQ-CXR.
  • Figure 3: Qualitative comparison of low-quality CXR synthesis from high-quality CXRs using bicubic interpolation, MUNIT-LQ, and DDPM-LQ.