Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Autopet III challenge: Incorporating anatomical knowledge into nnUNet for lesion segmentation in PET/CT

Hamza Kalisch, Fabian Hörst, Ken Herrmann, Jens Kleesiek, Constantin Seibold

TL;DR

A classifier is developed that identifies the tracer of the given PET/CT based on the Maximum Intensity Projection of the PET scan, and trained two individual nnUNet-ensembles for each tracer where anatomical labels are included as a multi-label task to enhance the model's performance.

Abstract

Lesion segmentation in PET/CT imaging is essential for precise tumor characterization, which supports personalized treatment planning and enhances diagnostic precision in oncology. However, accurate manual segmentation of lesions is time-consuming and prone to inter-observer variability. Given the rising demand and clinical use of PET/CT, automated segmentation methods, particularly deep-learning-based approaches, have become increasingly more relevant. The autoPET III Challenge focuses on advancing automated segmentation of tumor lesions in PET/CT images in a multitracer multicenter setting, addressing the clinical need for quantitative, robust, and generalizable solutions. Building on previous challenges, the third iteration of the autoPET challenge introduces a more diverse dataset featuring two different tracers (FDG and PSMA) from two clinical centers. To this extent, we developed a classifier that identifies the tracer of the given PET/CT based on the Maximum Intensity Projection of the PET scan. We trained two individual nnUNet-ensembles for each tracer where anatomical labels are included as a multi-label task to enhance the model's performance. Our final submission achieves cross-validation Dice scores of 76.90% and 61.33% for the publicly available FDG and PSMA datasets, respectively. The code is available at https://github.com/hakal104/autoPETIII/ .

Autopet III challenge: Incorporating anatomical knowledge into nnUNet for lesion segmentation in PET/CT

TL;DR

A classifier is developed that identifies the tracer of the given PET/CT based on the Maximum Intensity Projection of the PET scan, and trained two individual nnUNet-ensembles for each tracer where anatomical labels are included as a multi-label task to enhance the model's performance.

Abstract

Lesion segmentation in PET/CT imaging is essential for precise tumor characterization, which supports personalized treatment planning and enhances diagnostic precision in oncology. However, accurate manual segmentation of lesions is time-consuming and prone to inter-observer variability. Given the rising demand and clinical use of PET/CT, automated segmentation methods, particularly deep-learning-based approaches, have become increasingly more relevant. The autoPET III Challenge focuses on advancing automated segmentation of tumor lesions in PET/CT images in a multitracer multicenter setting, addressing the clinical need for quantitative, robust, and generalizable solutions. Building on previous challenges, the third iteration of the autoPET challenge introduces a more diverse dataset featuring two different tracers (FDG and PSMA) from two clinical centers. To this extent, we developed a classifier that identifies the tracer of the given PET/CT based on the Maximum Intensity Projection of the PET scan. We trained two individual nnUNet-ensembles for each tracer where anatomical labels are included as a multi-label task to enhance the model's performance. Our final submission achieves cross-validation Dice scores of 76.90% and 61.33% for the publicly available FDG and PSMA datasets, respectively. The code is available at https://github.com/hakal104/autoPETIII/ .
Paper Structure (20 sections, 2 figures, 6 tables)

This paper contains 20 sections, 2 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Pipeline overview
  4. MIP-based tracer classification
  5. Anatomy-guided lesion segmentation
  6. Experimental setup and evaluation
  7. Datasets
  8. Evaluation protocol and metrics
  9. Tracer classification.
  10. Lesion segmentation.
  11. Preprocessing and implementation details
  12. Tracer classification.
  13. Lesion segmentation.
  14. Results
  15. Baseline models
  16. ...and 5 more sections

Figures (2)

  • Figure 1: Pipeline of our final submission. In the first step, the PET volume from a given PET/CT scan is input into a classification module that determines the tracer type. PET and CT volumes are then concatenated and fed into a nnU-Net ensemble that was trained on the dataset of the classified tracer. The final binary lesion mask is output after postprocessing.
  • Figure 2: Exemplary coronal and sagittal MIPs for PET scans of two healthy patients using the FDG tracer and PSMA tracer. Whiter regions represent higher SUV values. In both views, the FDG MIP highlights clear higher uptake patterns for the brain and urinary bladder, whereas the latter reveals higher uptake in the kidneys, submandibular glands, and parotid glands.