Related Experiment Video
Updated: Jun 17, 2025
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
Impact of training data composition on the generalizability of convolutional neural network aortic cross-section
Chiara Manini1, Markus Hüllebrand2, Lars Walczak2
1Deutsches Herzzentrum der Charité (DHZC), Institute of Computer-assisted Cardiovascular Medicine, Berlin, Germany; Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt Universität zu Berlin, Berlin, Germany.
Insights
Deep learning models for four-dimensional cardiovascular magnetic resonance flow imaging (4D flow CMR) segmentation require diverse training data. Including various pathologies and technical factors improves model generalizability for accurate aortic hemodynamics analysis.
Area of Science:
- Cardiovascular Imaging
- Medical Image Analysis
- Deep Learning in Medicine
Background:
- Four-dimensional cardiovascular magnetic resonance flow imaging (4D flow CMR) is crucial for cardiovascular disease assessment.
- Manual segmentation of 4D flow data introduces variability, limiting reproducibility in hemodynamic analysis.
Purpose of the Study:
- To develop deep learning models for automatic segmentation in 4D flow CMR.
- To analyze the impact of training data characteristics on model generalization across diverse clinical settings.
Main Methods:
- Trained an enhanced 3D U-net CNN on 260 4D flow CMR datasets with varied pathologies and acquisition parameters.
- Evaluated model performance using Dice score, Hausdorff distance, and surface distance metrics.
- Compared automated segmentation-derived flow parameters with manual segmentation using Bland-Altman analysis.
Main Results:
- Training data representation of technical factors (scanner vendor, field strength) significantly influenced segmentation performance.
- Age impacted performance more than gender; bicuspid aortic valve (BAV) models generalized better to healthy subjects than vice versa.
- Heterogeneous training datasets are crucial for robust model generalization.
Conclusions:
- Diverse datasets encompassing various pathologies and technical acquisition factors are essential for training widely applicable automatic CNN segmentation models in 4D flow CMR.
- This approach enhances the reliability of hemodynamic visualization and quantitative parameter computation.
Background:
Four-dimensional cardiovascular magnetic resonance flow imaging (4D flow CMR) plays an important role in assessing cardiovascular diseases. However, the manual or semi-automatic segmentation of aortic vessel boundaries in 4D flow data introduces variability and limits the reproducibility of aortic hemodynamics visualization and quantitative flow-related parameter computation. This paper explores the potential of deep learning to improve 4D flow CMR segmentation by developing models for automatic segmentation and analyzes the impact of the training data on the generalization of the model across different sites, scanner vendors, sequences, and pathologies.
Methods:
The study population consists of 260 4D flow CMR datasets, including subjects without known aortic pathology, healthy volunteers, and patients with bicuspid aortic valve (BAV) examined at different hospitals. The dataset was split to train segmentation models on subsets with different representations of characteristics, such as pathology, gender, age, scanner model, vendor, and field strength. An enhanced three-dimensional U-net convolutional neural network (CNN) architecture with residual units was trained for time-resolved two-dimensional aortic cross-sectional segmentation. Model performance was evaluated using Dice score, Hausdorff distance, and average symmetric surface distance on test data, datasets with characteristics not represented in the training set (model-specific), and an overall evaluation set. Standard diagnostic flow parameters were computed and compared with manual segmentation results using Bland-Altman analysis and interclass correlation.
Results:
The representation of technical factors, such as scanner vendor and field strength, in the training dataset had the strongest influence on the overall segmentation performance. Age had a greater impact than gender. Models solely trained on BAV patients' datasets performed well on datasets of healthy subjects but not vice versa.
Conclusion:
This study highlights the importance of considering a heterogeneous dataset for the training of widely applicable automatic CNN segmentations in 4D flow CMR, with a particular focus on the inclusion of different pathologies and technical aspects of data acquisition.
