Personalizing Exercise Rehabilitation in Congenital Heart Disease: The Emerging Role of Artificial Intelligence

Background/Objective: Exercise training can improve fitness in patients with congenital heart disease (CHD), but differences in anatomy, cardiac function, and exercise responses, complicate intensity prescription.In this review, exercise prescription, ventilatory threshold assessment, and the potential application of artificial intelligence (AI)-assisted threshold assessment in patients with moderate to severe CHD, are examined.Design: Focused narrative review.Methods: Exercise interventions in patients with moderate to severe CHD, studies of ventilatory threshold validity and agreement, and methods for detecting or predicting thresholds from cardiopulmonary exercise testing (CPET), electrocardiography, or clinical data, were reviewed.The focused narrative search was conducted from August 2 to August 28, 2026.Results: Exercise programs used measured peak heart rate, heart rate reserve, perceived exertion, workload, or CPET thresholds, to prescribe intensity.Cardiovascular rehabilitation guidance favors ventilatory thresholds when reliable CPET data are available, although comparative benefits over other intensity methods are uncertain in patients with CHD.Threshold assessment varies with test procedures, signal quality, and assessor experience.Output from deep-learning models have shown agreement with expert assessments in healthy and mixed cardiovascular disease cohorts.Models using electrocardiographic or clinical data can also estimate threshold timing or heart rate.Regression equations can also estimate thresholds, although some require peak exercise measurements.Prospective validation of AI-derived estimates for exercise prescription for patients with moderate to severe CHD is limited.Conclusions: AI could support more consistent, clinician-confirmed threshold assessment.Before AI-derived estimates inform exercise prescription for patients with CHD, studies in which external validation of the models, quantification of errors in prescribed intensity, and evaluation of whether clinician-confirmed use is safe and clinically useful, should be performed.