Related Experiment Video
Updated: May 29, 2026
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Fully automatic left ventricle segmentation in [Formula: see text]Rb PET/CT Using a semi-supervised nnU-net
Mohammadreza Amirian1,2, Arthur Chevalley1,2, María Martín Asiain1,2
1Department of Nuclear Medicine and Molecular Imaging, Centre Hospitalier Universitaire Vaudois (CHUV), Lausanne, Switzerland.
Insights
A novel deep learning method automates left ventricle segmentation for Rubidium-82 PET/CT scans. This fully automatic pipeline improves accuracy and efficiency in myocardial blood flow quantification.
Area of Science:
- Cardiovascular Imaging
- Artificial Intelligence in Medicine
- Nuclear Cardiology
Background:
- Accurate quantification of myocardial blood flow (MBF) using Rubidium-82 (Rb-82) PET/CT necessitates precise left ventricle (LV) delineation.
- Manual or semi-automated LV segmentation is laborious and prone to errors, especially in areas of hypoperfused myocardium.
Purpose of the Study:
- To develop and validate a fully automatic LV segmentation pipeline using nnU-Net for Rb-82 PET/CT.
- To compare the performance of the automated pipeline against a semi-automatic thresholding baseline.
Main Methods:
- A manual, multimodal segmentation protocol was established as ground truth using dynamic PET and CT data from 40 patients.
- The nnU-Net model was trained using cross-validation and incorporated 805 additional unlabeled dynamic PET series via semi-supervised learning.
- Performance was evaluated against an optimized semi-automatic thresholding method.
Main Results:
- The nnU-Net pipeline significantly outperformed the baseline, achieving higher Dice scores (87.8% vs 75.1%), recall (89.1% vs 82.6%), and precision (88.1% vs 70.2%).
- Improvements were most significant in hypoperfused regions, with recall increasing by 20-30% compared to thresholding.
- Semi-supervised learning contributed to enhanced model robustness across rest and stress acquisitions.
Conclusions:
- A deep learning approach achieves fully automatic LV segmentation in Rb-82 PET/CT with accuracy comparable to expert performance.
- This automated framework eliminates manual segmentation, facilitating large-scale MBF quantification.
- The method supports reproducible, high-throughput cardiac PET analysis in clinical and research settings.
Background:
Quantification of myocardial blood flow (MBF) with [Formula: see text]Rb PET/CT requires accurate delineation of the left ventricle (LV). Manual or semi-automated contouring remains time-consuming and error-prone, particularly in hypoperfused myocardium. We developed and validated a fully automatic LV segmentation pipeline using nnU-Net applied to [Formula: see text]Rb PET/CT. A manual, multimodal segmentation protocol integrating dynamic PET and CT was established in a single-center cohort of 40 non-gated PET/CT series (20 patients, rest & stress), including challenging cases with extensive necrosis (median 21%). The resulting ground truth masks were used for five-fold cross-validation, and semi-supervised learning incorporated 805 additional unlabeled dynamic PET series (504 patients). Model performance was compared with an optimized semi-automatic thresholding baseline (35% [Formula: see text]).
Results:
The nnU-Net significantly outperformed the baseline, achieving a mean Dice of 87.8[85.6, 89.2]% vs 75.1[72.9, 76.9]%, recall 89.1[86.1, 91.4]% vs 82.6[79.1, 85.4]%, and precision 88.1[84.2, 90.4]% vs 70.2[67.2, 73.0]%. The improvement was most pronounced in hypoperfused regions, where recall increased by 20-30% compared to thresholding. Semi-supervised learning modestly enhanced model robustness across both rest and stress acquisitions.
Conclusions:
A deep-learning-based approach enables fully automatic LV segmentation in [Formula: see text]Rb PET/CT with near-expert accuracy. This framework eliminates manual interaction, supports large-scale MBF quantification, and paves the way for reproducible, high-throughput cardiac PET analysis in clinical and research workflows.