Artificial intelligence, extended reality and computational modelling in cross-sectional cardiovascular imaging in congenital heart disease: a narrative review

Introduction

Overview

Cross-sectional cardiovascular imaging, including computed tomography (CT) and cardiovascular magnetic resonance (CMR), is central to the comprehensive evaluation of children and adults with congenital heart disease (CHD) throughout their lifetime, from postnatal diagnosis to long-term follow-up and pre-interventional planning (1-4). The anatomical complexity and wide phenotypic variability of CHD, as well as the frequent need for surgical or catheter-based modifications, place high demands on image acquisition, post-processing, interpretation and clinical integration.

In recent years, advanced digital technologies, including artificial intelligence (AI), extended reality (XR), computational modelling, and three-dimensional (3D) printing, have increasingly been applied to CT- and CMR-based datasets to address these challenges (3-6). Rather than acting as isolated tools, these technologies are progressively forming integrated pipelines that span the entire imaging workflow, from acquisition and reconstruction to anatomical assessment, haemodynamic analysis, procedural planning, and prognostication (5,6).

Advanced technologies in CHD imaging

AI has rapidly evolved from semi-automated post-processing applications to more comprehensive frameworks encompassing image reconstruction, segmentation, automated quantification, phenotype recognition, and decision support (2,3,7). In the context of CHD, AI offers the potential to reduce observer variability, improve reproducibility, and enable scalable analysis of complex anatomies across heterogeneous patient populations (5,8,9).

XR is an umbrella term encompassing augmented reality (AR), virtual reality (VR), and mixed reality (MR), all of which enable immersive visualisation of 3D cardiovascular anatomy derived from cross-sectional imaging (10,11). In CHD, XR technologies enhance spatial understanding of complex intracardiac and vascular relationships and are increasingly used for surgical and interventional planning, multidisciplinary case discussion, medical education, and patient or family counselling (6,10-23).

Computational modelling, including computational fluid dynamics (CFD), leverages patient-specific anatomical reconstructions from CT and CMR to simulate blood flow, pressure distribution, wall shear stress, and device-vessel interactions (20,24,25). These models provide functional insights that complement anatomical imaging and support hypothesis testing, risk stratification, and optimisation of surgical or catheter-based strategies, particularly in complex circulations such as Fontan physiology or reconstructed right ventricular outflow tracts (15,24,26).

Closely related to computational modelling, additive manufacturing, most commonly 3D printing, enables the creation of physical, patient-specific cardiovascular models derived from cross-sectional imaging data (20,27-32). These models have demonstrated value in improving anatomical comprehension, facilitating procedural rehearsal, supporting device selection, and enhancing communication among clinicians, trainees, patients, and families (30-35). Increasingly, 3D printing is integrated with computational simulation workflows, linking virtual haemodynamic analysis with tangible physical models (24,25,34).

The integration of cross-sectional imaging with computational modelling, XR, and 3D printing workflows is summarised in Figure 1.

Figure 1 Integration of cross-sectional imaging with computational modelling, extended reality, and 3D printing. 3D, three-dimensional; 4D, four-dimensional; CMR, cardiovascular magnetic resonance; CT, computed tomography.

Rationale and objectives

Despite rapid technological progress, the application of AI, XR, computational modelling and 3D printing in CHD is still hindered by a lack of data, anatomical and surgical heterogeneity, limited external validation and incomplete integration into routine clinical workflows (9,10,36). Regulatory, ethical, and medico-legal considerations further influence clinical adoption, particularly in paediatric and lifelong congenital care (3,9,29-31).

The objective of this narrative review is to provide a structured overview of current clinical applications and research developments related to AI, XR, computational modelling, and 3D printing based on CT and CMR imaging in CHD. By synthesising existing evidence, highlighting strengths and limitations, and discussing emerging integrated pipelines, this review aims to offer a coherent framework for understanding the present state of the field and to outline future directions for clinical translation (5,6).

The review is structured to first introduce AI applications in cross-sectional imaging for CHD, followed by sections on XR and computational modelling, before addressing workflow integration, limitations, and future perspectives.

The literature search was updated to October 2025 to include recent methodological advances and preliminary clinical translation studies. We present this article in accordance with the Narrative Review reporting checklist (available at https://cdt.amegroups.com/article/view/10.21037/cdt-2025-1-616/rc).

Methods

This manuscript is a narrative review aimed at synthesising current developments in AI, XR, and computational modelling applied to cross-sectional cardiovascular imaging in CHD. A targeted literature search was performed in PubMed to identify representative and clinically relevant publications published between January 2014 and October 2025, including early online and ahead-of-print articles.

The search strategy was designed to capture key methodological advances, clinical applications, and translational challenges rather than to provide exhaustive coverage. Given the narrative nature of this review, no formal screening or selection process consistent with scoping or systematic reviews was undertaken. Articles were selected based on relevance to the review objectives, methodological contribution, and clinical applicability. Case reports and non-English publications were excluded.

To enhance transparency, Table 1 summarises the databases searched, key search terms, and overall search approach used to inform this narrative synthesis.

AI

Cross-sectional imaging is central to the diagnosis, risk stratification and long-term surveillance of CHD. However, the interpretation of complex CHD and their anatomical differences from structurally normal cardiac anatomy, as well as their surgical modifications and physiology, poses challenges to reproducibility and inter-operator consistency in image interpretation. Over the past decade, AI research has evolved from post-processing applications to integrated frameworks encompassing acquisition acceleration, anatomical assessment, computational flow measurements, XR applications and clinical decision support.

An overview of the main AI paradigms and their applications in cross-sectional imaging for CHD is illustrated in Figure 2.

Figure 2 Artificial intelligence applications across the cross-sectional imaging workflow in congenital heart disease. 2D, two-dimensional; 3D, three-dimensional; 4D, four-dimensional; CNN, convolutional neural network.

AI encompasses a range of computational approaches that enable automated pattern recognition, classification, and prediction from imaging data. In cross-sectional cardiovascular imaging, AI methods can broadly be divided into traditional machine-learning techniques and deep-learning approaches (2,3,8). Traditional machine learning relies on handcrafted features derived from imaging data and has been applied to tasks such as disease classification, anomaly detection, and outcome prediction in CHD, but its performance is highly dependent on expert-driven feature selection and limited scalability (9,10,36). In contrast, deep learning (DL), most commonly implemented using convolutional neural networks (CNNs), learns hierarchical features directly from raw images and currently represents the dominant paradigm in cross-sectional imaging (3,8,37). Two-dimensional (2D) CNNs are widely used for segmentation, tissue characterisation, and functional assessment on slice-based datasets, while 3D CNNs exploit volumetric information and are increasingly applied to whole-heart segmentation, ventricular volumetry, and flow or deformation-based analyses in CHD (38-40). Other AI approaches, such as natural language processing, are primarily used to extract structured information from imaging reports, facilitate dataset curation, and support automated reporting and decision-support pipelines (9,41,42). A major limitation across all AI approaches is the requirement for large, high-quality training datasets: traditional machine-learning models may be trained on hundreds to a few thousand cases, whereas robust deep-learning models, particularly 3D architectures, often require several thousand to tens of thousands of annotated studies to achieve generalisable performance (8,37,40). The need for expert manual annotation, coupled with variability between centres in imaging protocols, scanner vendors and surgical anatomy, remains a key barrier to scalability and clinical translation, particularly in paediatric and congenital populations where data availability is inherently limited (9,10).

Along this trajectory, frameworks focusing on CHD and paediatric-to-adult continuity of care have proposed a ‘teach-predict-plan-guide’ imaging continuum. This continuum positions AI not merely as an efficient tool, but as a coordinator of personalised CHD care (5,6,8).

In this framework, the “teach” component refers to AI-supported visualisation and anatomical understanding, including automated segmentation, 3D reconstructions, and educational tools that enhance clinical interpretation of complex congenital anatomy. The “predict” component encompasses AI-driven quantitative analysis and modelling aimed at forecasting functional outcomes, disease progression, or procedural risk based on imaging-derived features. The “plan” stage relates to the use of AI outputs to support pre-procedural decision-making, including selection and optimisation of surgical or interventional strategies, often in combination with computational modelling or XR tools. Finally, the “guide” component represents the integration of AI-assisted imaging into procedural guidance, follow-up assessment, and longitudinal care, enabling personalised and adaptive management across the lifespan of patients with CHD.

Where available, findings from prospective and validation studies are prioritised; however, the relative lack of randomised controlled trials reflects the early translational stage of these technologies in CHD.

Evolution of AI in cross-sectional imaging for CHD

Early AI applications (~2015–2018) focused on semi-automated ventricular volume measurements and simple segmentation algorithms adapted from acquired heart disease datasets (38,42-45). During the period from 2019 to 2022, the first studies on DL-based whole-heart segmentation and reconstruction in CHD patients were developed and adapted for patients with a Fontan circulation and repaired tetralogy of Fallot. Recent developments (since 2023) have introduced publicly available CHD imaging datasets and multi-stage AI pipelines that can feed segmentation outputs into CFD, XR holography and simulation environments used for cardiovascular surgery (27,39,46).

Milestones regarding AI in cross-sectional cardiovascular imaging are shown in Table 2 (8,10,43,47-58).

Acquisition and reconstruction

DL-based reconstruction techniques can reduce scan times and mitigate motion and noise artefacts by learning directly from raw k-space data. This is particularly advantageous for paediatric patients or those who are unable to cooperate, as breath-holding can be difficult for them and they may require sedation or general anaesthesia. The 2024 release of the CMRxRecon dataset introduced an open-access and multimodal k-space resource framework. This enables fair comparisons and makes DL-based reconstruction pipelines more reproducible. Although these datasets are not specific to CHD, they can be easily adapted for CHD protocols via transfer learning (30).

AI-assisted acceleration and denoising of 4D Flow CMR have been shown to enhance the stability of wall shear stress and flow metrics. These improved data provide a more reliable and structured basis for subsequent computational haemodynamic modelling (37,59) which can be clinically important in patients with aortopathies or Fontan circulation.

Anatomy and function assessment

CMR segmentation

DL-models now demonstrate accurate cardiac segmentation across heterogeneous CHD anatomies, addressing long-standing issues of labour-intensive contouring and high inter-observer variability. In 2024, Pace et al. proposed standardized guidelines to promote generalizable architectures, and to ensure fair comparison between segmentation methods (40).

The HVSMR-2.0 dataset further addresses a major gap by releasing expert-labelled CHD CMR volumes with annotations for all four chambers and great vessels, including mandatory and optional vessel extents to reflect true anatomic variability and surgical repair patterns (40). This dataset can be used to train neural networks to segment cardiovascular structures in CHD patients, as well as for other AI applications.

From a clinical adoption perspective, AI-based cardiac segmentation represents one of the most mature applications in cross-sectional imaging. Several segmentation and automated quantification tools are already integrated into commercial CMR and CT software platforms, facilitating routine clinical use for volumetric and functional assessment. In contrast, more advanced applications that rely on downstream interpretation of segmented data, such as automated phenotype recognition or prognostic modelling, remain largely confined to research settings and pilot studies with limited external validation.

CT segmentation and derived measurements

AI-based CT segmentation is still a challenge. A general overview in complex CHD emphasized the need for labelling strategies that capture anatomical variants (10). Although current pipelines are methodologically heterogeneous, they consistently demonstrate feasibility and are moving towards more standardised validation frameworks with explicit mesh quality checks to ensure usability for surgical planning and holographic navigation.

Automated measurements of volumes, mass and vessel dimensions

Automated extraction of chamber volumes, myocardial mass and vessel diameters from CMR and CT scans have substantially reduced manual processing time while improving reproducibility, which is useful for monitoring CHD, including ventricular function assessments and measurement of aortic dimensions. Beyond individual metric extraction, end-to-end interpretation models are emerging as a potential inflection point. A large-scale CMR-based screening and diagnostic framework trained on 9,719 patients is now available. Although it was not developed for CHD, it provides a compelling blueprint for CHD-specific retraining, particularly for CMR in complex anatomies (60,61).

Phenotype recognition, anomaly detection, and decision support

Anomaly detection and decision-support tools are increasingly being adapted for use in CHD-specific workflows (41,62-64). A recent article provides a structured overview of the integration of AI systems across multimodal imaging pipelines, from automated feature extraction to risk stratification and longitudinal follow-up, while emphasizing the importance of embedding outputs directly into clinical data streams (e.g., electrocardiograms and electronic health records) (42). Reviews in this area report improvements in the diagnostic performance of echocardiography, CMR and CT, especially with regard to modelling perioperative risk prediction. However, despite these advances, most models remain inadequately tested in lesion-specific cohorts. Conditions such as tetralogy of Fallot, transposition of the great arteries, double outlet right ventricle, post-coarctation repair and conduit/baffle anatomies require external validation strategies that reflect the morphological heterogeneity observed in clinical practice (10,60). These strategies must also include explicit mechanisms for failure detection and clinician override.

Collectively, these studies demonstrate that AI-based approaches can improve diagnostic consistency and risk stratification in selected CHD populations; however, most evidence derives from retrospective or single-centre cohorts, underscoring the need for lesion-specific external validation.

Despite promising results, most AI-based phenotype recognition and decision-support tools in CHD remain at a research or pilot stage. These models are predominantly derived from retrospective or single-centre datasets and are not yet widely embedded in clinical workflows. As such, their likelihood of near-term clinical adoption depends on lesion-specific external validation, regulatory approval, and seamless integration into existing reporting and decision-making environments.

XR

In CHD, XR technologies are primarily applied to enhance anatomical understanding, support surgical and interventional planning, and facilitate education and communication. XR is an umbrella term for technologies that create computer-generated environments or objects and includes AR, VR and MR (11,12). The various technologies are defined by the relationship between the real and virtual worlds. While users perceive virtual objects as an extension of the real world in AR, they are immersed in a purely virtual world with VR. Another essential feature of XR is that all forms of it are immersive technologies.

Applications

XR has a wide range of applications in medicine, and there is growing clinical and research interest in various areas of cardiology (11,12). XR can improve the visualisation of the intracardiac and the depth perception and hereby the understanding of the underlying cardiac anatomy (60). Several publications have shown its usefulness for medical education and surgical as well as interventional planning based on cross-sectional imaging data (15,17-23,65-67). Majority of VR tools need a previous segmentation of cardiac structure, while more recent technologies are able to produce virtual models directly from 3D data sets, avoiding potential bias of the intermediate step of segmentation (68,69). Thus, Tandon et al. and Stephenson et al. showed the usefulness of VR planning for the interventional closure of sinus venosus defects (67,68,70). Whereas others showed the usefulness of XR for the planning of transcatheter pulmonary valve replacements (71).

Across the available literature, XR applications consistently improve anatomical understanding and procedural planning in complex CHD; however, evidence is predominantly derived from feasibility studies and prospective observational cohorts rather than randomised comparisons (35-38,72,73).

In terms of clinical adoption, XR technologies are currently most mature in applications related to education, anatomical visualisation, and pre-procedural planning, where they enhance understanding without directly altering clinical decision pathways. Conversely, intra-procedural guidance and real-time XR navigation remain largely investigational, with current evidence primarily derived from feasibility studies and small prospective cohorts (74).

Computational modelling

Principles of computational modelling in CHD

The emergence of high-resolution cross-sectional cardiovascular imaging, particularly volumetric CT and CMR, has transformed the evaluation and management of CHD. These advanced imaging techniques allow detailed, patient-specific reconstructions of cardiac anatomy and provide clinicians with an unprecedented view of complex structural relationships that underlie cardiac congenital malformations (75).

Cross-sectional imaging techniques yield 3D data sets that capture complex spatial relationships of cardiac chambers, outflow tracts, vessels and congenital defects (76). With the use of these datasets, computational modelling frameworks already may generate virtual representations of flow dynamics, wall shear stress and device interactions, forming a bridge from anatomy to function (20,24).

CFD and haemodynamic simulation

These computational models use advanced segmentation and meshing techniques to simulate blood flow and tissue mechanics. This helps reveal the influence of anatomical abnormalities or surgical repairs on post-operative cardiovascular function (25,77).

3D printing and physical modelling

Meanwhile, additive manufacturing, specifically, the 3D printing of patient-specific cardiovascular models, has evolved into a valuable clinical tool. 3D models derived from cross-sectional imaging data provide improved visual and spatial comprehension compared to traditional 2D or virtual renderings (Figure 3) (28-32). In addition to their educational value and their role in facilitating communication with families, these models allow surgeons and interventionalists to rehearse procedures, select appropriately sized devices and construct physical flow setups for bench-top haemodynamic testing (20,33).

Figure 3 3D modelling using CMR data. 3D model of a patient with a double outlet right ventricle from different directions (A-C,E,F). In the other views the blood pool (D) and an interior view of the model (G) is shown. 3D, three-dimensional; CMR, cardiovascular magnetic resonance.

Integration of modelling and 3D printing

The synergy of computational modelling with 3D printing closes the loop from imaging to simulation to physical artefact. A typical workflow begins with segmentation of cross-sectional imaging data, generation of a virtual geometric model, meshing for CFD or finite element analysis, and ultimately fabrication of a tangible model that incorporates insights from the simulation (34). This integrated approach enables pre-interventional evaluation of flow patterns (e.g., residual shunts, vortex formation), device-vessel interaction, and even the manufacturing of models with patient-specific compliance or tissue properties for surgical rehearsal (24,25,34). Recent reviews emphasise that for structural heart disease, including CHD, such workflows are already shifting training paradigms and procedural planning (35). Together, computational modelling and 3D printing show promise in personalising CHD treatment, reducing intra-operative surprises and improving outcomes for this challenging patient group.

Evidence synthesis and current limitations

Prospective and simulation-based studies suggest that computational modelling can provide clinically relevant haemodynamic insights and support procedural planning; however, widespread adoption is currently limited by computational complexity, modelling assumptions, and the lack of outcome-driven prospective trials.

Although computational modelling and patient-specific simulation can provide clinically meaningful haemodynamic insights, their routine clinical adoption remains limited. Factors such as computational complexity, modelling assumptions, processing time, and the need for specialised expertise currently restrict these approaches to selected complex cases or research-driven workflows rather than widespread clinical implementation.

Workflow integration, evaluation, and reporting

Recent open-access resources now provide a foundation for more reproducible CHD imaging pipelines (40,78). However, their effectiveness depends on shifting the focus from generic performance metrics to lesion-stratified evaluations, multi-site generalisation testing and formal uncertainty quantification. In practice, human-in-the-loop quality control remains essential, particularly in the presence of metallic implants, stents or flow-related artefacts that may compromise automated outputs. Ideally, segmentation-derived structures (center lines, volumes and flow maps) should feed directly into structured reporting environments and enable the export of watertight surface meshes that are suitable for XR visualisation and computational flow dynamics. This would close the loop between imaging, procedural planning and follow-up.

Limitations and challenges

There are several challenges to the use of AI, XR and computational modelling in cross-sectional imaging in CHD. Key barriers in include data scarcity, anatomical and surgical heterogeneity, different age groups in pediatric patients and variations across cross-sectional imaging protocols. This results in a lack of standardization (78) as well as large multicenter trials that can show improved clinical outcomes.

In addition, legal regulations vary between countries, and ethical concerns have been raised and addressed (60-62). However, given the growing use of AI, XR and modelling, updates to medico-legal regulations and their harmonization are expected to be necessary.

CFD-pipelines are still slow, require modelling assumptions and extensive technical experience (79). Together, this significantly limits the use of CFD in the current clinical practice. Regarding XR there is a lack of standard for 3D model fidelity in comparison standard imaging methods. Furthermore, the technical requirements are only available in some centers, and this currently limits the widespread use of XR in clinical scenarios such as joint surgical conferences (12).

Finally, there is a lack of widely available learning frameworks for patients with CHD, where follow-up cross-sectional imaging continuously refines AI models via federated or privacy-preserving learning.

From an economic perspective, the cost-benefit balance of AI, XR, and computational modelling in CHD remains heterogeneous and highly dependent on institutional scale and infrastructure. At present, hospitals are more likely to adopt commercial AI solutions that are embedded within existing CT and CMR platforms, as these tools require limited additional hardware investment and offer immediate efficiency gains through reduced post-processing time and improved reproducibility. In contrast, advanced applications such as patient-specific computational modelling, XR planning, or large-scale 3D model production entail higher upfront costs related to specialised software licenses, hardware, data storage, and expert personnel for segmentation, model generation, and data analysis.

The production of large-scale patient-specific models, whether virtual or physical, remains resource-intensive, requiring dedicated imaging post-processing, quality-controlled segmentation, and, in the case of 3D printing, material and manufacturing costs. As a result, these approaches are currently best justified for selected complex cases where the expected clinical or educational benefit outweighs the associated expense. Looking forward, increasing automation, standardisation of pipelines, federated learning strategies, and broader commercial integration are expected to reduce marginal costs and lower barriers to adoption. In parallel, hospitals’ decisions to invest in these technologies will increasingly depend on demonstrable improvements in workflow efficiency, clinical outcomes, and training value rather than technical feasibility alone.

Future outlook

It is reasonable to suggest that AI-based image reconstruction, segmentation and quantitative reporting using CMR and CT will advance further in the near future in CHD patients (3,9,71,80). Other advances will likely improve image-based CFD, such as GPU-accelerated CFD, and will make patient-specific haemodynamic quantification more practical for planning CHD therapy and risk stratification (30). Furthermore, XR will likely move forward from proof-of-concept and education to real clinical planning and procedural guidance (26,79).

Conclusions

AI is no longer confined to post-hoc quantification. It is now shaping the entire imaging-to-decision continuum and will continue to develop further in the coming years. This is also true for patients with CHD, covering acquisition, reconstruction, phenotype recognition, risk stratification, automated reporting, and intervention planning. However, the realization of this potential of AI will depend on multicenter data curation, carefully stratified validation frameworks and governance models that reflect the unique demands of lifelong congenital care.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the Narrative Review reporting checklist. Available at https://cdt.amegroups.com/article/view/10.21037/cdt-2025-1-616/rc

Peer Review File: Available at https://cdt.amegroups.com/article/view/10.21037/cdt-2025-1-616/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://cdt.amegroups.com/article/view/10.21037/cdt-2025-1-616/coif). I.V. serves as an unpaid editorial board member of Cardiovascular Diagnosis and Therapy from February 2026 to December 2027. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Ponzoni M, Bertelli F, Yoo SJ, et al. Mixed reality for preoperative planning and intraoperative assistance of surgical correction of complex congenital heart defects. J Thorac Cardiovasc Surg 2025;170:327-335.e1. [Crossref] [PubMed]
2. Fotaki A, Puyol-Antón E, Chiribiri A, et al. Artificial Intelligence in Cardiac MRI: Is Clinical Adoption Forthcoming? Front Cardiovasc Med 2021;8:818765. [Crossref] [PubMed]
3. Mastrodicasa D, van Assen M, Huisman M, et al. Use of AI in Cardiac CT and MRI: A Scientific Statement from the ESCR, EuSoMII, NASCI, SCCT, SCMR, SIIM, and RSNA. Radiology 2025;314:e240516. [Crossref] [PubMed]
4. Ghosh RM, Jolley MA, Mascio CE, et al. Clinical 3D modeling to guide pediatric cardiothoracic surgery and intervention using 3D printed anatomic models, computer aided design and virtual reality. 3D Print Med 2022;8:11.
5. Salavitabar A, Figueroa CA, Lu JC, et al. Emerging 3D technologies and applications within congenital heart disease: teach, predict, plan and guide. Future Cardiol 2020;16:695-709. [Crossref] [PubMed]
6. Chessa M, Van De Bruaene A, Farooqi K, et al. Three-dimensional printing, holograms, computational modelling, and artificial intelligence for adult congenital heart disease care: an exciting future. Eur Heart J 2022;43:2672-84. [PubMed]
7. Blonsky SE, Henning RE, Masterson EM, et al. Virtual Reality Curriculum Increases Pediatric Clerkship Students' Knowledge of Congenital Heart Disease. Pediatr Cardiol 2026;47:552-6. [Crossref] [PubMed]
8. Sethi Y, Patel N, Kaka N, et al. Artificial Intelligence in Pediatric Cardiology: A Scoping Review. J Clin Med 2022;11:7072. [Crossref] [PubMed]
9. Pozza A, Zanella L, Castaldi B, et al. How Will Artificial Intelligence Shape the Future of Decision-Making in Congenital Heart Disease? J Clin Med 2024;13:2996. [Crossref] [PubMed]
10. Jone PN, Gearhart A, Lei H, et al. Artificial Intelligence in Congenital Heart Disease: Current State and Prospects. JACC Adv 2022;1:100153. [Crossref] [PubMed]
11. Gupta SK, Gupta P. Anaglyph stereo virtual dissection: a novel inexpensive method for stereoscopic visualisation of intracardiac anatomy on CT angiogram. Cardiol Young 2021;31:1958-61. [Crossref] [PubMed]
12. Lau I, Gupta A, Ihdayhid A, et al. Clinical Applications of Mixed Reality and 3D Printing in Congenital Heart Disease. Biomolecules 2022;12:1548. [Crossref] [PubMed]
13. d'Aiello AF, Cabitza F, Natali C, et al. The Effect of Holographic Heart Models and Mixed Reality for Anatomy Learning in Congenital Heart Disease: An Exploratory Study. J Med Syst 2023;47:64. [Crossref] [PubMed]
14. da Silva MJN, de Souza Soares AM, Braga HFGM, et al. Transforming Understanding of Congenital Heart Disease Through a Virtual Reality Innovation. Comput Inform Nurs 2025;43:e01266. [Crossref] [PubMed]
15. Daungsupawong H, Wiwanitkit V. Deep Learning-Based Method for Rapid 3D Whole-Heart Modeling in Congenital Heart Disease: Correspondence. Cardiology 2025;150:451-2. [Crossref] [PubMed]
16. Kim B, Loke YH, Mass P, et al. A Novel Virtual Reality Medical Image Display System for Group Discussions of Congenital Heart Disease: Development and Usability Testing. JMIR Cardio 2020;4:e20633. [Crossref] [PubMed]
17. Gehrsitz P, Rompel O, Schöber M, et al. Cinematic Rendering in Mixed-Reality Holograms: A New 3D Preoperative Planning Tool in Pediatric Heart Surgery. Front Cardiovasc Med 2021;8:633611. [Crossref] [PubMed]
18. Goo HW, Park SJ, Yoo SJ. Advanced Medical Use of Three-Dimensional Imaging in Congenital Heart Disease: Augmented Reality, Mixed Reality, Virtual Reality, and Three-Dimensional Printing. Korean J Radiol 2020;21:133-45. [Crossref] [PubMed]
19. Galeczka M, Smerdzinski S, Tyc F, et al. Virtual reality for transcatheter procedure planning in congenital heart disease. Kardiol Pol 2023;81:1026-7. [Crossref] [PubMed]
20. Huang H, Wu Y. A Deep Learning-Based Method for Rapid 3D Whole-Heart Modeling in Congenital Heart Disease. Cardiology 2025;150:243-58. [Crossref] [PubMed]
21. Milano EG, Kostolny M, Pajaziti E, et al. Enhanced 3D visualization for planning biventricular repair of double outlet right ventricle: a pilot study on the advantages of virtual reality. Eur Heart J Digit Health 2021;2:667-75. [Crossref] [PubMed]
22. Bertelli F, Raimondi F, Godard C, et al. Fast-track virtual reality software to facilitate 3-dimensional reconstruction in congenital heart disease. Interdiscip Cardiovasc Thorac Surg 2023;36:ivad087. [Crossref] [PubMed]
23. Raimondi F, Vida V, Godard C, et al. Fast-track virtual reality for cardiac imaging in congenital heart disease. J Card Surg 2021;36:2598-602. [Crossref] [PubMed]
24. Sun Z. Patient-Specific 3D-Printed Models in Pediatric Congenital Heart Disease. Children (Basel) 2023;10:319. [Crossref] [PubMed]
25. Yahiro DS, Cruz MP, Ribeiro BFC, et al. Impact of 3D Printing on Cardiac Surgery in Congenital Heart Diseases: A Systematic Review and Meta-Analysis. Arq Bras Cardiol 2024;121:e20240430. [PubMed]
26. Agarwal A, Ranard L, Vahl T, et al. Advanced Computer Simulation Based on Cardiac Imaging in Planning of Structural Heart Disease Interventions. J Clin Med 2025;14:6885. [Crossref] [PubMed]
27. Cantinotti M, Valverde I, Kutty S. Three-dimensional printed models in congenital heart disease. Int J Cardiovasc Imaging 2017;33:137-44. [Crossref] [PubMed]
28. Anwar S, Singh GK, Miller J, et al. 3D Printing is a Transformative Technology in Congenital Heart Disease. JACC Basic Transl Sci 2018;3:294-312. [Crossref] [PubMed]
29. Ghosh RM, Thompson M. Tetralogy of Fallot With Pulmonary Atresia and Major Aortopulmonary Collateral Arteries: Diagnostic Modalities-The Role of Computed Tomography, Cardiac Magnetic Resonance Imaging, and Three-Dimensional Modeling. World J Pediatr Congenit Heart Surg 2025;16:191-6. [Crossref] [PubMed]
30. Spanaki A, Kabir S, Stephenson N, et al. 3D Approaches in Complex CHD: Where Are We? Funny Printing and Beautiful Images, or a Useful Tool? J Cardiovasc Dev Dis 2022;9:269. [Crossref] [PubMed]
31. Anagnostopoulos S, Baltayiannis N, Koletsis NE, et al. 3D printing in medicine: bridging imaging, education, and practice. Arch Med Sci Atheroscler Dis 2025;10:e172-e188. [Crossref] [PubMed]
32. Bertelli F, Singh A, Lam CZ, et al. Advancements in 3D modeling technologies for congenital heart disease: integrating 3D printing, virtual reality, and holograms. Pediatr Radiol 2026;56:28-39. [Crossref] [PubMed]
33. Wang DD, Qian Z, Vukicevic M, et al. 3D Printing, Computational Modeling, and Artificial Intelligence for Structural Heart Disease. JACC Cardiovasc Imaging 2021;14:41-60. [Crossref] [PubMed]
34. Schafstedde M, Yevtushenko P, Nordmeyer S, et al. Virtual treatment planning in three patients with univentricular physiology using computational fluid dynamics-Pitfalls and strategies. Front Cardiovasc Med 2022;9:898701. [Crossref] [PubMed]
35. Haupt M, Maurer MH, Thomas RP. Explainable Artificial Intelligence in Radiological Cardiovascular Imaging-A Systematic Review. Diagnostics (Basel) 2025;15:1399. [Crossref] [PubMed]
36. Mohammadi I, Rajai Firouzabadi S, Hosseinpour M, et al. Using artificial intelligence to predict post-operative outcomes in congenital heart surgeries: a systematic review. BMC Cardiovasc Disord 2024;24:718. [Crossref] [PubMed]
37. Tolu-Akinnawo OZ, Ezekwueme F, Omolayo O, et al. Advancements in Artificial Intelligence in Noninvasive Cardiac Imaging: A Comprehensive Review. Clin Cardiol 2025;48:e70087. [Crossref] [PubMed]
38. Winther HB, Hundt C, Schmidt B, et al. ν-net: Deep Learning for Generalized Biventricular Mass and Function Parameters Using Multicenter Cardiac MRI Data. JACC Cardiovasc Imaging 2018;11:1036-8. [Crossref] [PubMed]
39. Ankenbrand MJ, Lohr D, Schlötelburg W, et al. Deep learning-based cardiac cine segmentation: Transfer learning application to 7T ultrahigh-field MRI. Magn Reson Med 2021;86:2179-91. [Crossref] [PubMed]
40. Pace DF, Contreras HTM, Romanowicz J, et al. HVSMR-2.0: A 3D cardiovascular MR dataset for whole-heart segmentation in congenital heart disease. Sci Data 2024;11:721. [Crossref] [PubMed]
41. Antoun I, Nizam A, Ebeid A, et al. Artificial Intelligence in Adult Congenital Heart Disease: Diagnostic and Therapeutic Applications and Future Directions. Rev Cardiovasc Med 2025;26:41523. [Crossref] [PubMed]
42. Alnasser TN, Abdulaal L, Maiter A, et al. Advancements in cardiac structures segmentation: a comprehensive systematic review of deep learning in CT imaging. Front Cardiovasc Med 2024;11:1323461. [Crossref] [PubMed]
43. Queirós S, Barbosa D, Engvall J, et al. Multi-centre validation of an automatic algorithm for fast 4D myocardial segmentation in cine CMR datasets. Eur Heart J Cardiovasc Imaging 2016;17:1118-27. [Crossref] [PubMed]
44. Shaaf ZF, Jamil MMA, Ambar R, et al. Automatic Left Ventricle Segmentation from Short-Axis Cardiac MRI Images Based on Fully Convolutional Neural Network. Diagnostics (Basel) 2022;12:414. [Crossref] [PubMed]
45. Avendi MR, Kheradvar A, Jafarkhani H. A combined deep-learning and deformable-model approach to fully automatic segmentation of the left ventricle in cardiac MRI. Med Image Anal 2016;30:108-19. [Crossref] [PubMed]
46. Sophocleous F, Delchev K, De Garate E, et al. Feasibility of Wave Intensity Analysis from 4D Cardiovascular Magnetic Resonance Imaging Data. Bioengineering (Basel) 2023;10:662. [Crossref] [PubMed]
47. Shahzad R, Tao Q, Dzyubachyk O, et al. Fully-automatic left ventricular segmentation from long-axis cardiac cine MR scans. Med Image Anal 2017;39:44-55. [Crossref] [PubMed]
48. Krasnobaev A, Sozykin A. An Overview of Techniques for Cardiac Left Ventricle Segmentation on Short-Axis MRI. ITM Web Conf 2016;8:01003.
49. Ma C, Luo G, Wang K. A Combined Random Forests and Active Contour Model Approach for Fully Automatic Segmentation of the Left Atrium in Volumetric MRI. Biomed Res Int 2017;2017:8381094. [Crossref] [PubMed]
50. Pace DF, Dalca AV, Geva T, et al. Interactive Whole-Heart Segmentation in Congenital Heart Disease. Med Image Comput Comput Assist Interv 2015;9351:80-8. [Crossref] [PubMed]
51. Karimi-Bidhendi S, Arafati A, Cheng AL, et al. Fully-automated deep-learning segmentation of pediatric cardiovascular magnetic resonance of patients with complex congenital heart diseases. J Cardiovasc Magn Reson 2020;22:80. [Crossref] [PubMed]
52. Fathi MF, Perez-Raya I, Baghaie A, et al. Super-resolution and denoising of 4D-Flow MRI using physics-Informed deep neural nets. Comput Methods Programs Biomed 2020;197:105729. [Crossref] [PubMed]
53. Ferdian E, Suinesiaputra A, Dubowitz DJ, et al. 4DFlowNet: Super-Resolution 4D Flow MRI Using Deep Learning and Computational Fluid Dynamics. Front Phys 2020;8:138. [Crossref]
54. Park S, Chung M. Cardiac segmentation on CT Images through shape-aware contour attentions. Comput Biol Med 2022;147:105782. [Crossref] [PubMed]
55. Keramati H, Lu X, Cabanag M, et al. Applications and advances of immersive technology in cardiology. Curr Probl Cardiol 2024;49:102762. [Crossref] [PubMed]
56. Silva A, Vale N. Digital Twins in Personalized Medicine: Bridging Innovation and Clinical Reality. J Pers Med 2025;15:503. [Crossref] [PubMed]
57. Wang C, Lyu J, Wang S, et al. CMRxRecon: A publicly available k-space dataset and benchmark to advance deep learning for cardiac MRI. Sci Data 2024;11:687. [Crossref] [PubMed]
58. Strocchi M, Hammersley DJ, Halliday BP, et al. Cardiac digital twins: a tool to investigate the function and treatment of the diabetic heart. Cardiovasc Diabetol 2025;24:293. [Crossref] [PubMed]
59. Kim B, Nguyen P, Loke YH, et al. Virtual Reality Cardiac Surgical Planning Software (CorFix) for Designing Patient-Specific Vascular Grafts: Development and Pilot Usability Study. JMIR Cardio 2022;6:e35488. [Crossref] [PubMed]
60. Wang YJ, Yang K, Wen Y, et al. Screening and diagnosis of cardiovascular disease using artificial intelligence-enabled cardiac magnetic resonance imaging. Nat Med 2024;30:1471-80. [Crossref] [PubMed]
61. Gabbert DD, Petersen L, Burleigh A, et al. Detection of hypoplastic left heart syndrome anatomy from cardiovascular magnetic resonance images using machine learning. MAGMA 2024;37:115-25. [Crossref] [PubMed]
62. Diller GP, Orwat S, Vahle J, et al. Prediction of prognosis in patients with tetralogy of Fallot based on deep learning imaging analysis. Heart 2020;106:1007-14. [Crossref] [PubMed]
63. Shiri I, Baj G, Mohammadi Kazaj P, et al. AI-based detection and classification of anomalous aortic origin of coronary arteries using coronary CT angiography images. Nat Commun 2025;16:3095. [Crossref] [PubMed]
64. Xu X, Jia Q, Yuan H, et al. A clinically applicable AI system for diagnosis of congenital heart diseases based on computed tomography images. Med Image Anal 2023;90:102953. [Crossref] [PubMed]
65. Benesch Vidal ML, Arvanitaki A, Diller GP. The role of artificial intelligence and mobile health in diagnosis and management of pulmonary arterial hypertension. Int J Cardiol Congenit Heart Dis 2025;22:100622. [Crossref] [PubMed]
66. Ismail Sharieff I, Ravikumar DB, Joshi S, et al. Applications of augmented reality in cardiology till 2024: a comprehensive review of innovations and clinical impacts. Front Virtual Real 2025;6:1580619. [Crossref]
67. Stephenson N, Rosenthal E, Jones M, et al. Virtual Reality for Preprocedure Planning of Covered Stent Correction of Superior Sinus Venosus Atrial Septal Defects. Circ Cardiovasc Interv 2024;17:e013964. [Crossref] [PubMed]
68. Tandon A, Burkhardt BEU, Batsis M, et al. Sinus Venosus Defects: Anatomic Variants and Transcatheter Closure Feasibility Using Virtual Reality Planning. JACC Cardiovasc Imaging 2019;12:921-4. [Crossref] [PubMed]
69. Jolley MA, Lasso A, Nam HH, et al. Toward predictive modeling of catheter-based pulmonary valve replacement into native right ventricular outflow tracts. Catheter Cardiovasc Interv 2019;93:E143-52. [Crossref] [PubMed]
70. Chen C, Qin C, Qiu H, et al. Deep Learning for Cardiac Image Segmentation: A Review. Front Cardiovasc Med 2020;7:25. [Crossref] [PubMed]
71. Roshanfar M, Salimi M, Jang SJ, et al. Emerging Image-Guided Navigation Techniques for Cardiovascular Interventions: A Scoping Review. Bioengineering (Basel) 2025;12:488. [Crossref] [PubMed]
72. Kanschik D, Bruno RR, van Genderen ME, et al. Extended reality in cardiovascular care: a systematic review. Eur Heart J Digit Health 2025;6:878-87. [Crossref] [PubMed]
73. Stephenson N, Pushparajah K, Wheeler G, et al. Extended reality for procedural planning and guidance in structural heart disease - a review of the state-of-the-art. Int J Cardiovasc Imaging 2023;39:1405-19. [Crossref] [PubMed]
74. Gerrah R, Haller SJ. Computational fluid dynamics: a primer for congenital heart disease clinicians. Asian Cardiovasc Thorac Ann 2020;28:520-32. [Crossref] [PubMed]
75. Dave A, Santos RD, Siddiqi U, et al. Applications of Computational Fluid Dynamics in Congenital Heart Disease: A Review. J Cardiovasc Dev Dis 2025;12:70. [Crossref] [PubMed]
76. Frieberg P, Sjöberg P, Hedström E, et al. In vivo hepatic flow distribution by computational fluid dynamics can predict pulmonary flow distribution in patients with Fontan circulation. Sci Rep 2023;13:18206. [Crossref] [PubMed]
77. Auger DA, Zhong X, Epstein FH, et al. Semi-automated left ventricular segmentation based on a guide point model approach for 3D cine DENSE cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2014;16:8. [Crossref] [PubMed]
78. Tandon A, Byrne N, Nieves Velasco Forte Mde L, et al. Use of a semi-automated cardiac segmentation tool improves reproducibility and speed of segmentation of contaminated right heart magnetic resonance angiography. Int J Cardiovasc Imaging 2016;32:1273-9. [Crossref] [PubMed]
79. Leivaditis V, Beltsios E, Papatriantafyllou A, et al. Artificial Intelligence in Cardiac Surgery: Transforming Outcomes and Shaping the Future. Clin Pract 2025;15:17. [Crossref] [PubMed]
80. Sachdeva R, Armstrong AK, Arnaout R, et al. Novel Techniques in Imaging Congenital Heart Disease: JACC Scientific Statement. J Am Coll Cardiol 2024;83:63-81. [Crossref] [PubMed]