Anomaly-focused Single Image Super-resolution with Artifact Removal for Chest X-rays using Distribution-aware Diffusion Model

Dattatreyo Roy, Angshuman Paul
Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, PMLR 250:1297-1309, 2024.

Abstract

Single image super-resolution (SISR) is a crucial task in the field of medical imaging. It transforms low-resolution images into high-resolution counterparts. Performing SISR on chest x-ray images enhances image quality, aiding better diagnosis. However, artifacts may be present in the images. We propose an anomaly-guided SISR process utilizing the denoising mechanism of the diffusion model to iteratively remove noise and restore the original image. We train the model to learn the data distribution, enabling it to eliminate artifacts within the images. Additionally, we ensure reconstruction of the disease regions by prioritizing their reconstruction. Our research experiment over the publicly available dataset and find that the existing SISR methods are unable to learn and remove these artificially added artifacts. On the other hand, our proposed model not only prioritizes superior image reconstruction but also remove the artifacts. Our method is found to outperform the existing methods. The code is publicly available at https://github.com/Datta-IITJ/MIDL_code.git.

Cite this Paper

BibTeX
@InProceedings{pmlr-v250-roy24a, title = {Anomaly-focused Single Image Super-resolution with Artifact Removal for Chest X-rays using Distribution-aware Diffusion Model}, author = {Roy, Dattatreyo and Paul, Angshuman}, booktitle = {Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning}, pages = {1297--1309}, year = {2024}, editor = {Burgos, Ninon and Petitjean, Caroline and Vakalopoulou, Maria and Christodoulidis, Stergios and Coupe, Pierrick and Delingette, Hervé and Lartizien, Carole and Mateus, Diana}, volume = {250}, series = {Proceedings of Machine Learning Research}, month = {03--05 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v250/main/assets/roy24a/roy24a.pdf}, url = {https://proceedings.mlr.press/v250/roy24a.html}, abstract = {Single image super-resolution (SISR) is a crucial task in the field of medical imaging. It transforms low-resolution images into high-resolution counterparts. Performing SISR on chest x-ray images enhances image quality, aiding better diagnosis. However, artifacts may be present in the images. We propose an anomaly-guided SISR process utilizing the denoising mechanism of the diffusion model to iteratively remove noise and restore the original image. We train the model to learn the data distribution, enabling it to eliminate artifacts within the images. Additionally, we ensure reconstruction of the disease regions by prioritizing their reconstruction. Our research experiment over the publicly available dataset and find that the existing SISR methods are unable to learn and remove these artificially added artifacts. On the other hand, our proposed model not only prioritizes superior image reconstruction but also remove the artifacts. Our method is found to outperform the existing methods. The code is publicly available at https://github.com/Datta-IITJ/MIDL_code.git.} }
Endnote
%0 Conference Paper %T Anomaly-focused Single Image Super-resolution with Artifact Removal for Chest X-rays using Distribution-aware Diffusion Model %A Dattatreyo Roy %A Angshuman Paul %B Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2024 %E Ninon Burgos %E Caroline Petitjean %E Maria Vakalopoulou %E Stergios Christodoulidis %E Pierrick Coupe %E Hervé Delingette %E Carole Lartizien %E Diana Mateus %F pmlr-v250-roy24a %I PMLR %P 1297--1309 %U https://proceedings.mlr.press/v250/roy24a.html %V 250 %X Single image super-resolution (SISR) is a crucial task in the field of medical imaging. It transforms low-resolution images into high-resolution counterparts. Performing SISR on chest x-ray images enhances image quality, aiding better diagnosis. However, artifacts may be present in the images. We propose an anomaly-guided SISR process utilizing the denoising mechanism of the diffusion model to iteratively remove noise and restore the original image. We train the model to learn the data distribution, enabling it to eliminate artifacts within the images. Additionally, we ensure reconstruction of the disease regions by prioritizing their reconstruction. Our research experiment over the publicly available dataset and find that the existing SISR methods are unable to learn and remove these artificially added artifacts. On the other hand, our proposed model not only prioritizes superior image reconstruction but also remove the artifacts. Our method is found to outperform the existing methods. The code is publicly available at https://github.com/Datta-IITJ/MIDL_code.git.
APA
Roy, D. & Paul, A.. (2024). Anomaly-focused Single Image Super-resolution with Artifact Removal for Chest X-rays using Distribution-aware Diffusion Model. Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 250:1297-1309 Available from https://proceedings.mlr.press/v250/roy24a.html.

Related Material