Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AutoPET Challenge: Tumour Synthesis for Data Augmentation

Lap Yan Lennon Chan, Chenxin Li, Yixuan Yuan

TL;DR

The potential of leveraging the deep prior from a generative model to serve as a data augmenter for automated lesion segmentation in PET/CT scans is explored, potentially enhancing the accuracy and reliability of cancer diagnostics.

Abstract

Accurate lesion segmentation in whole-body PET/CT scans is crucial for cancer diagnosis and treatment planning, but limited datasets often hinder the performance of automated segmentation models. In this paper, we explore the potential of leveraging the deep prior from a generative model to serve as a data augmenter for automated lesion segmentation in PET/CT scans. We adapt the DiffTumor method, originally designed for CT images, to generate synthetic PET-CT images with lesions. Our approach trains the generative model on the AutoPET dataset and uses it to expand the training data. We then compare the performance of segmentation models trained on the original and augmented datasets. Our findings show that the model trained on the augmented dataset achieves a higher Dice score, demonstrating the potential of our data augmentation approach. In a nutshell, this work presents a promising direction for improving lesion segmentation in whole-body PET/CT scans with limited datasets, potentially enhancing the accuracy and reliability of cancer diagnostics.

AutoPET Challenge: Tumour Synthesis for Data Augmentation

TL;DR

The potential of leveraging the deep prior from a generative model to serve as a data augmenter for automated lesion segmentation in PET/CT scans is explored, potentially enhancing the accuracy and reliability of cancer diagnostics.

Abstract

Accurate lesion segmentation in whole-body PET/CT scans is crucial for cancer diagnosis and treatment planning, but limited datasets often hinder the performance of automated segmentation models. In this paper, we explore the potential of leveraging the deep prior from a generative model to serve as a data augmenter for automated lesion segmentation in PET/CT scans. We adapt the DiffTumor method, originally designed for CT images, to generate synthetic PET-CT images with lesions. Our approach trains the generative model on the AutoPET dataset and uses it to expand the training data. We then compare the performance of segmentation models trained on the original and augmented datasets. Our findings show that the model trained on the augmented dataset achieves a higher Dice score, demonstrating the potential of our data augmentation approach. In a nutshell, this work presents a promising direction for improving lesion segmentation in whole-body PET/CT scans with limited datasets, potentially enhancing the accuracy and reliability of cancer diagnostics.
Paper Structure (18 sections, 5 figures, 2 tables)

This paper contains 18 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Exploratory Data Analysis
  3. Insufficient dataset size
  4. Countermeasures
  5. Related Work
  6. Towards Generalizable Tumour Synthesis (DiffTumor) difftumour
  7. Methodology
  8. DynUnet baseline
  9. DiffTumor adaptation
  10. Results
  11. Experimental Setting
  12. Quantitative Result
  13. Qualitative Results
  14. Predicted Mask
  15. Synthesized Tumour
  16. ...and 3 more sections

Figures (5)

  • Figure 1: Dice score for tumour segmentation models trained with different numbers of cases (real or synthesized), the trend shows that in general, more training cases lead to better performance. The result of DiffTumor difftumour on 11k is re-implemented.
  • Figure 2: The successful and unsuccessful case of lesion segmentation
  • Figure 3: More Lesion Segmentation Results
  • Figure 4: More Lesion Segmentation Results
  • Figure 5: Tumour Generation in unseen data. The generation network learns well by not automatically translating every tumour to be shown as a PET tracer as that is not always the case.