Asztalos, Ivor

Purpose of review Artificial intelligence applied to electrocardiography (AI-ECG) has rapidly been investigated in adult cardiovascular medicine, yet translation into pediatric and congenital heart disease populations has lagged. This review summarizes contemporary AI-ECG methodologies and emerging applications in pediatric and congenital heart disease (PCHD), with emphasis on current clinical utility, technical challenges, and future opportunities for implementation. Recent findings Recent studies demonstrate that deep learning models can accurately identify arrhythmias, ventricular dysfunction, and CHD from standard ECG. Convolutional neural networks remain the dominant architecture, although transformer-based foundation models and self-supervised learning approaches are increasingly being explored. AI-ECG applications in PCHD have expanded from automated interpretation toward proactive risk stratification, including prediction of ventricular dysfunction, mortality, and sudden cardiac death risk. Additional work has investigated wearable monitoring, telemetry analysis, and integration with longitudinal clinical data. Despite promising performance, most studies remain retrospective and single-center, with limited external validation and challenges related to small datasets, physiologic heterogeneity, and age-dependent ECG variation. Summary AI-ECG has the potential to transform PCHD by improving diagnostic accuracy, enabling earlier disease detection, and enhancing longitudinal risk assessment. Broader clinical implementation will require multicenter collaboration, prospective validation, standardized datasets, and careful attention to ethical, regulatory, and equity considerations.

ABSTRACT Background Coronary artery (CA) injury from catheter ablation is an uncommon complication, but the risk is under‐recognized. The risk for CA injury significantly increases when a CA is within 3 mm of the site of ablation, but CA imaging to assess for CA proximity to ablation sites is not commonly performed in children prior to ablation. Objective The aim of the study was to assess CA proximity to potential ablation sites. Methods A retrospective review of all coronary computed tomography angiograms (CTAs) in pediatric patients <18 years of age at Mayo Clinic from 2017 through 2023 was conducted. The proximity of CAs to potential ablation sites in the coronary sinus and the appendage bases were measured. Results Seventy‐two CTAs in 67 patients >15 kg and <18 years old were included. Median age was 12 years, and median weight was 41 kg. In right dominant CA circulation, 23/36 (64%) had a CA within 3 mm of the coronary sinus. In left and co‐dominant CA circulation, 26/26 (100%) had a CA within 3 mm of the coronary sinus. The right CA was within 3 mm of the appendage base in 49/69 (71%). The circumflex CA was within 3 mm of the appendage base in 50/68 (74%). Conclusion Most pediatric patients have a CA ≤ 3 mm from potential ablation sites in the coronary sinus and base of the appendages. Precautions should be taken to avoid CA damage before performing catheter ablation in these areas.

Abstract Aims To assess the potential for artificial intelligence-enabled electrocardiogram (AI-ECG) to serve as a long-term cardiac surveillance tool and predict left ventricular systolic dysfunction in childhood cancer patients. Methods and results We assessed performance of our previously established AI-ECG model to predict left ventricular ejection fraction (LVEF) ≤50% and ≤40% in patients with childhood cancer during internal testing (Boston Children’s Hospital) and external validation (Children’s Hospital of Philadelphia). The internal test cohort comprised 447 patients [57% male; age at cancer diagnosis 11.2 (5.4–15.7) years; 1553 ECG–echo pairs at median age 13.5 (IQR 7.7–17.9) years; 6.4% with LVEF ≤50%; 1.3% with LVEF ≤40%], 28% with leukaemia, 16% with lymphoma, 8% with neuroblastoma, 8% with sarcoma, 2% with gastrointestinal cancers, 3% with genitourinary cancers, 6% with central nervous system cancers, 11% with other/unspecified cancers, and 18% with missing/unknown cancer labels. Treatment strategies included anthracyclines (35%), bone marrow transplant (7%), and radiation (1%). The external test cohort comprised 2964 patients [55.4% male; 7054 ECG–echo pairs at median age 11.6 (IQR 6.8–15.1) years; 2.5% with LVEF ≤50%; 0.9% with LVEF ≤40%]. Similar AUROCs (0.80–0.85), sensitivities (0.75–0.82), NPVs (0.986–0.996), and percent predicted negative (51–65%) were obtained across institutions to predict LVEF ≤50%, outperforming a biomarker-based model benchmark. Patients with high AI-ECG risk scores for LVEF ≤50% had higher rates of mortality [hazard ratio 3.1 (95% CI 1.8–5.3), P &lt; 0.001] compared to patients with low AI-ECG risk scores. Conclusion AI-ECG shows promise as a digital biomarker for cardiac surveillance in the vulnerable childhood cancer survivor population.

Abstract Background Most incidental gastric polyps identified during upper endoscopy are considered low-risk. However, current guidelines recommend sampling all gastric polyps for histopathologic analysis. We aimed to devise a simple narrow-band imaging (NBI) classification to reduce the need for routine biopsies of low-risk gastric polyps. Methods Pairs of NBI and white-light images were collected from 73 gastric polyps for which concurrent histopathologic diagnosis was available. A diagnostic accuracy cohort study was performed. Two blinded endoscopists independently analysed NBI features of each polyp for color, vessel pattern, surface pattern, and any combinations thereof to develop a classification scheme to differentiate low-risk polyps (fundic-gland or hyperplastic) from high-risk polyps (adenomatous or adenocarcinoma) and fundic-gland polyps (FGPs) from non-FGPs. Results An isolated lacy vessel pattern and a homogenous absence of surface pattern successfully differentiated low-risk from high-risk gastric polyps. Combining both descriptors into a single algorithm resulted in a negative predictive value (NPV) of 100% [95% confidence interval (CI): 100%–100%], positive predictive value (PPV) of 13.7% (95% CI: 2.6–24.8), sensitivity of 100% (95% CI: 100%–100%), and specificity of 53.7% (95% CI: 45.3%–62.0%) for high-risk polyps. This would reduce the number of polyps requiring biopsy by 50%, while still capturing all high-risk polyps. Regarding FGPs, using a rule not to biopsy polyps with isolated lacy vessels resulted in a 94.9% NPV (95% CI: 89.2%–100%), 63.2% PPV (95% CI: 47.2%–79.2%), 94.8% sensitivity (95% CI: 89.5%–100%), and 63.6% specificity (95% CI: 51.3%–76.0%) for non-FGPs. Conclusion In this derivation cohort study, NBI is helpful for differentiating between high-risk and low-risk gastric polyps, thereby reducing the need for routine sampling of low-risk polyps. These results need to be validated in a separate test population.