Title : Pixels to practice: AI integrated 3D printing of patient specific congenital heart models for cardiac surgical planning and training: A literature review
Abstract:
Aim: Three-dimensional (3D) printing has transformed surgical planning and education around congenital heart disease (CHD) by using patient-specific anatomical visualisation. However, manual image segmentation is a major bottleneck and limits scalability. This review explores the advances of introducing Artificial Intelligence (AI) to automate cardiac reconstruction facilitating 3D printing, augmented reality (AR) and virtual reality (VR) applications.
Methodology: A focused search (Pubmed, Scopus 2015-2025) identified 6 original studies that used AI-based segmentation or reconstruction for 3D CHD model generation. Extracted data included imaging modality, dataset size, AI algorithm type, segmentation accuracy (dice score), processing time and reported clinical or education outcomes. Supporting clinical studies for 3D model generation were analysed to contextualise cost and feasibility.
Results: AI integration using deep learning models (U-Net) has been successfully tested across CT, MRI and 3D rotational angiography (3DRA) imaging modalities. The AI models achieved Dice scores of 0.78-0.93, reported improved accuracy by ~10-12% and reduced segmentation time from 6-10 hours to about 30 minutes. Graph-matching techniques enabled vessel classification in anatomically complex hearts, providing accurate 3D reconstructions. These models were successfully adapted for 3D printing, AR and VR visualization and were validated for anatomical fidelity.
Conclusion: AI-integrated segmentation or reconstruction significantly accelerates and enhances 3D model generation for CHD across different imaging modalities, improving accessibility of patient-specific 3D model generation. Integration of deep-learning with graph-based anatomy shows strong potential to revolutionize preoperative planning and surgical training. However, multicentre validation with larger cohort sizes is essential to confirm the scalability, reproducibility and clinical impact.
