Bone supression in planar X-ray images with Stable Diffusion

Daniel Sanderson, Mariana Elizalde, Paula Ochotorena, Manuel Desco, Monica Abella

Keywords: X-ray, radiography, bone supression, diffusion models, Stable Diffusion

TL;DR: This paper explores the applicability of foundational models such as Stable Diffusion to bone supression in X-ray images.

Abstract: Bone suppression is a processing technique that aims to enhance the visualization of chest radiographic images by attenuating bones while preserving soft tissue details. This has been achieved with deep learning methods but they either introduce blurring or do not fully remove bones. In this work, we propose a bone removal method for radiography based on Stable Diffusion. To address the lack of publicly available bone suppression datasets, the model is pre-trained using a synthetic dataset simulated from computed tomography scans. Preliminary evaluation demonstrates the ability of the proposed model to fully remove bones while preserving spatial resolution.

Submission Number: 28