Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Quality Enhancement of Radiographic X-ray Images by Interpretable Mapping

Hongxu Yang, Najib Akram Aboobacker, Xiaomeng Dong, German Gonzalez, Lehel Ferenczi, Gopal Avinash

TL;DR

This work tackles the problem of inconsistent X-ray image presentation caused by varied acquisition conditions. It introduces an interpretable deep-learning framework that predicts per-pixel parameter maps to drive regional and global LUT-based brightness/contrast adjustments, yielding explainable enhancement maps. The model achieves competitive quantitative performance ($PSNR = 24.75$ dB, $SSIM = 0.843$) while maintaining regional detail and presentation consistency across anatomies, and it provides interpretable maps to understand the enhancement behavior. Practically, this approach can reduce radiologist workload by standardizing image appearance and offering transparent explanations of automatic augmentations, with potential extensions to semi-supervised or unsupervised learning on larger datasets.

Abstract

X-ray imaging is the most widely used medical imaging modality. However, in the common practice, inconsistency in the initial presentation of X-ray images is a common complaint by radiologists. Different patient positions, patient habitus and scanning protocols can lead to differences in image presentations, e.g., differences in brightness and contrast globally or regionally. To compensate for this, additional work will be executed by clinical experts to adjust the images to the desired presentation, which can be time-consuming. Existing deep-learning-based end-to-end solutions can automatically correct images with promising performances. Nevertheless, these methods are hard to be interpreted and difficult to be understood by clinical experts. In this manuscript, a novel interpretable mapping method by deep learning is proposed, which automatically enhances the image brightness and contrast globally and locally. Meanwhile, because the model is inspired by the workflow of the brightness and contrast manipulation, it can provide interpretable pixel maps for explaining the motivation of image enhancement. The experiment on the clinical datasets show the proposed method can provide consistent brightness and contrast correction on X-ray images with accuracy of 24.75 dB PSNR and 0.8431 SSIM.

Quality Enhancement of Radiographic X-ray Images by Interpretable Mapping

TL;DR

This work tackles the problem of inconsistent X-ray image presentation caused by varied acquisition conditions. It introduces an interpretable deep-learning framework that predicts per-pixel parameter maps to drive regional and global LUT-based brightness/contrast adjustments, yielding explainable enhancement maps. The model achieves competitive quantitative performance ( dB, ) while maintaining regional detail and presentation consistency across anatomies, and it provides interpretable maps to understand the enhancement behavior. Practically, this approach can reduce radiologist workload by standardizing image appearance and offering transparent explanations of automatic augmentations, with potential extensions to semi-supervised or unsupervised learning on larger datasets.

Abstract

X-ray imaging is the most widely used medical imaging modality. However, in the common practice, inconsistency in the initial presentation of X-ray images is a common complaint by radiologists. Different patient positions, patient habitus and scanning protocols can lead to differences in image presentations, e.g., differences in brightness and contrast globally or regionally. To compensate for this, additional work will be executed by clinical experts to adjust the images to the desired presentation, which can be time-consuming. Existing deep-learning-based end-to-end solutions can automatically correct images with promising performances. Nevertheless, these methods are hard to be interpreted and difficult to be understood by clinical experts. In this manuscript, a novel interpretable mapping method by deep learning is proposed, which automatically enhances the image brightness and contrast globally and locally. Meanwhile, because the model is inspired by the workflow of the brightness and contrast manipulation, it can provide interpretable pixel maps for explaining the motivation of image enhancement. The experiment on the clinical datasets show the proposed method can provide consistent brightness and contrast correction on X-ray images with accuracy of 24.75 dB PSNR and 0.8431 SSIM.
Paper Structure (4 sections, 1 equation, 4 figures, 1 table)

This paper contains 4 sections, 1 equation, 4 figures, 1 table.

Table of Contents

  1. INTRODUCTION
  2. METHODS
  3. RESULTS
  4. CONCLUSIONS

Figures (4)

  • Figure 1: From raw image to left: global brightness and contrast adjustment for initial presentation after several iterations. From raw image to right: workflow for brightness and contrast enhancement at regional and global levels.
  • Figure 2: (a) Interpretable maps prediction and enhanced image generation based on input image, predicted maps (a, b, wc and ww) and remap formula. (b) Regional LUT and global LUT for dynamic range of [0,65535].
  • Figure 3: Top eight examples are selected from different models in Table 1, including the ground truth (GT). Bottom eight examples are enlarged ankle regions of corresponding model. The details of the bone structures are preserved by our method.
  • Figure 4: Top: Outputs from the proposed method, which are adjusted regionally and globally for brightness and contrast. Bottom: Automated brightness and contrast adjustment globally only.