Lungmend corrects lung masks on severe tuberculosis ct scans

LungMend: A Healthy-Only Latent Diffusion Shape Prior for Lung Mask Correction in Severe Tuberculosis CT

Computer Vision and Pattern Recognition

Summary

Lung scans of people with serious tuberculosis can be hard to analyze because the disease changes the lung’s shape in ways that confuse computers. The authors created LungMend, a tool that fixes these mistakes by learning what healthy lungs look like and then guessing the complete lung shape, even when parts are damaged or missing. Unlike previous methods, LungMend does this without needing examples of diseased lungs or the original scan details. This helps doctors see the full lung area more accurately in severe cases.

What this means in practice

  • For hospital radiology teams: Correct lung segmentation masks on CT scans affected by severe tuberculosis to improve diagnostic accuracy and treatment planning.
  • For medical imaging software developers: Integrate a healthy-lung shape prior to enhance lung segmentation tools for cases with pathological abnormalities without needing pathological training data.$Commercial implications: Enables development of improved lung segmentation software packages specialized for severe lung disease cases, enhancing marketable imaging products.

Authors

Rezkellah Noureddine Khiat, Pierre-Yves Brillet, Catalin Fetita

Abstract

Lung segmentation models such as R231 are widely used, but they fail on severe tuberculosis (TB) CT: cavities and consolidations look like non-lung tissue and are left out of the mask. We present LungMend, a post-processing model that corrects such masks using only healthy anatomy. A latent diffusion model, trained on healthy lungs with synthetically corrupted masks, learns to recover the complete lung from the faulty mask and the bone anatomy, without any pathological data or CT intensities. LungMend recovers cavities and consolidations that R231 misses and improves accuracy on expert-annotated slices. Code: https://github.com/noureddinekhiati/LungMend