Deep learning for cardiac CT segmentation for congenital heart disease: A systematic review
- 1Engineering for Health Research Centre, Aston University, Aston Triangle, Birmingham, United Kingdom
- 2Aston Institute of Health & Neurodevelopment, Aston University, Aston Triangle, Birmingham, United Kingdom
- 3School of Computer Science and Digital Technologies, Aston University, Aston Triangle, Birmingham, United Kingdom
Abstract
Purpose: Deep Learning (DL) has transformed cardiac image segmentation, yet its application to congenital heart disease (CHD) remains underexplored, with no prior systematic review in this emerging domain. This review evaluates current DL approaches for CHD CT segmentation, identifies best performing model architectures, assesses quality of reporting and highlights challenges for clinical implementation. Materials and methods: Following PRISMA guidelines, six databases: PubMed, IEEE Xplore, MDPI, Science Direct, Web of Science, and Scopus were searched (2004–2025). Extracted characteristics include source of dataset, ground truth labelling approach, DL architecture and evaluation methods. Reporting quality was evaluated using the Checklist for Artificial Intelligence in Medical imaging (CLAIM). Results: 14 studies met the inclusion criteria, reflecting the emergence of this research area. The UNet architecture and its variants dominated (8 studies). The highest reported Dice Similarity Coefficient (DSC) for cardiac structures segmentation was achieved utilising an advanced hybrid ResNet-based model (Aorta, DSC = 0.945). Well documented areas include the training approach (12 studies) and model description (all). Only two studies performed external validation, and two assessed inter- and intra-rater variability. No studies reported sample size determination and handing of missing data. Small vascular structures consistently underperformed compared to whole heart segmentation. Conclusion: Deep learning models show strong potential for CHD CT segmentation but remain limited by small datasets, inconsistent reporting and lack of clinical evaluation. Advancing this field requires multidisciplinary collaboration, standardised reporting, and integration of clinical co- design. As the first comprehensive review in this area, we provide a strong baseline for evaluating future studies.
Related
- Fetal cardiac interventions versus expectant management in congenital heart disease: a systematic review and meta-analysis.
- Risk factors for acute gastrointestinal injury after cardiac surgery in underweight children with congenital heart disease: a single-center retrospective study
- Symbolic transfer entropy reveals enhanced cardiac autonomic influence on cortical activity in infants with congenital heart disease
- Home-based neuro-cardiac rehabilitation in congenital heart disease: the QUALINEURO-REHAB trial study protocol
- Examining the Clinical Utility of Deep Learning Models for Perivascular Space Segmentation in Alzheimer's Disease
BibTeX
@article{yu2026deep,
title = {{Deep learning for cardiac CT segmentation for congenital heart disease: A systematic review}},
author = {Sinling Tiffany Yu and Zahra Amini and Shereen Fouad and Jan Novak and Antonio Fratini},
journal = {PLOS One},
year = {2026},
month = sep,
doi = {10.1371/journal.pone.0357600},
url = {https://doi.org/10.1371/journal.pone.0357600},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Deep learning for cardiac CT segmentation for congenital heart disease: A systematic review
AU - Sinling Tiffany Yu
AU - Zahra Amini
AU - Shereen Fouad
AU - Jan Novak
AU - Antonio Fratini
PY - 2026
DA - 2026/09/29
JO - PLOS One
DO - 10.1371/journal.pone.0357600
UR - https://doi.org/10.1371/journal.pone.0357600
AB - Purpose: Deep Learning (DL) has transformed cardiac image segmentation, yet its application to congenital heart disease (CHD) remains underexplored, with no prior systematic review in this emerging domain. This review evaluates current DL approaches for CHD CT segmentation, identifies best performing model architectures, assesses quality of reporting and highlights challenges for clinical implementation. Materials and methods: Following PRISMA guidelines, six databases: PubMed, IEEE Xplore, MDPI, Science Direct, Web of Science, and Scopus were searched (2004–2025). Extracted characteristics include source of dataset, ground truth labelling approach, DL architecture and evaluation methods. Reporting quality was evaluated using the Checklist for Artificial Intelligence in Medical imaging (CLAIM). Results: 14 studies met the inclusion criteria, reflecting the emergence of this research area. The UNet architecture and its variants dominated (8 studies). The highest reported Dice Similarity Coefficient (DSC) for cardiac structures segmentation was achieved utilising an advanced hybrid ResNet-based model (Aorta, DSC = 0.945). Well documented areas include the training approach (12 studies) and model description (all). Only two studies performed external validation, and two assessed inter- and intra-rater variability. No studies reported sample size determination and handing of missing data. Small vascular structures consistently underperformed compared to whole heart segmentation. Conclusion: Deep learning models show strong potential for CHD CT segmentation but remain limited by small datasets, inconsistent reporting and lack of clinical evaluation. Advancing this field requires multidisciplinary collaboration, standardised reporting, and integration of clinical co- design. As the first comprehensive review in this area, we provide a strong baseline for evaluating future studies.
SN - 1932-6203
ER -