Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Disentangled PET Lesion Segmentation

Tanya Gatsak, Kumar Abhishek, Hanene Ben Yedder, Saeid Asgari Taghanaki, Ghassan Hamarneh

TL;DR

3D disentanglement method that uses a 3D UNet-like encoder-decoder architecture to disentangle disease and normal healthy anatomical features with losses for segmentation, reconstruction, and healthy component plausibility is proposed.

Abstract

PET imaging is an invaluable tool in clinical settings as it captures the functional activity of both healthy anatomy and cancerous lesions. Developing automatic lesion segmentation methods for PET images is crucial since manual lesion segmentation is laborious and prone to inter- and intra-observer variability. We propose PET-Disentangler, a 3D disentanglement method that uses a 3D UNet-like encoder-decoder architecture to disentangle disease and normal healthy anatomical features with losses for segmentation, reconstruction, and healthy component plausibility. A critic network is used to encourage the healthy latent features to match the distribution of healthy samples and thus encourages these features to not contain any lesion-related features. Our quantitative results show that PET-Disentangler is less prone to incorrectly declaring healthy and high tracer uptake regions as cancerous lesions, since such uptake pattern would be assigned to the disentangled healthy component.

Disentangled PET Lesion Segmentation

TL;DR

3D disentanglement method that uses a 3D UNet-like encoder-decoder architecture to disentangle disease and normal healthy anatomical features with losses for segmentation, reconstruction, and healthy component plausibility is proposed.

Abstract

PET imaging is an invaluable tool in clinical settings as it captures the functional activity of both healthy anatomy and cancerous lesions. Developing automatic lesion segmentation methods for PET images is crucial since manual lesion segmentation is laborious and prone to inter- and intra-observer variability. We propose PET-Disentangler, a 3D disentanglement method that uses a 3D UNet-like encoder-decoder architecture to disentangle disease and normal healthy anatomical features with losses for segmentation, reconstruction, and healthy component plausibility. A critic network is used to encourage the healthy latent features to match the distribution of healthy samples and thus encourages these features to not contain any lesion-related features. Our quantitative results show that PET-Disentangler is less prone to incorrectly declaring healthy and high tracer uptake regions as cancerous lesions, since such uptake pattern would be assigned to the disentangled healthy component.

Paper Structure

This paper contains 12 sections, 6 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Method
  3. Modified Encoder Architecture
  4. Segmentation Decoder and Prediction
  5. Image Decoder and Reconstruction
  6. Critic Network for Healthy Distribution Matching
  7. Overall Objective Function
  8. Experiments
  9. Results
  10. Conclusions
  11. Compliance with Ethical Standards
  12. Acknowledgments

Figures (2)

  • Figure 1: The proposed disentanglement architecture of PET-Disentangler facilitates the learning of disease features through two pathways: a segmentation prediction path and image reconstruction path. By re-entangling healthy and disease-specific features, this design enables PET-Disentangler to effectively capture disease characteristics while maintaining an accurate representation of healthy anatomy. The black arrows indicate feature flow throughout the network, the blue arrows represent skip connections, and the green arrow represents the use of the mask prediction in the image decoder.
  • Figure 2: Comparison of PET-Disentangler to SegOnly, with red markers highlighting false positives for healthy uptake patterns produced by SegOnly, which are not present in PET-Disentangler.