Artificial intelligence - assisted individualized cardiac rehabilitation in heart failure: a prospective study
A K E Freitas, A B Oberderfer, B C Chiaratti, G C Mendes, G P Kakuda, M M Roldan, S M NetoAbstract
Introduction
Cardiac rehabilitation (CR) is a non-pharmacological intervention recommended by clinical guidelines for the management of heart failure (HF), being associated with improvements in functional capacity, quality of life, and reductions in hospitalizations. However, limited access, low adherence, and a shortage of specialized professionals remain significant barriers to its implementation. In this context, artificial intelligence (AI) emerges as a potential tool to individualize CR protocols, optimize functional gains, and expand access, particularly within public healthcare systems.
Objective
To evaluate the clinical and functional effectiveness of an AI–based individualized CR protocol compared with conventional rehabilitation and with the absence of structured exercise in patients with HF.
Methods
This was a prospective, and comparative study that included 15 patients with HF with reduced ejection fraction (mean EF 26.9%), New York Heart Association functional class I–III, clinically stable for at least 30 days. Participants were randomized into three parallel groups: Group A (AI-individualized CR, n = 6), Group B (conventional CR, n = 6), and Group C (no structured exercise, n = 3). The AI-based protocol was generated using a customized decision-support tool developed through a specific IA-bot, fed with CR guidelines, clinical data, and patients’ functional test results. Outcomes included functional capacity (6-minute walk test, sit-to-stand test, handgrip strength) and quality of life (Kansas City Cardiomyopathy Questionnaire – KCCQ), assessed over an 8-week period. Statistical analysis was performed according to data distribution, with significance set at p <0.05
Results
All groups undergoing CR showed clinical and functional improvements compared with baseline. Group A demonstrated greater gains in objective functional outcomes, particularly in 6-minute walk distance and handgrip strength, compared with Groups B and C. Conventional rehabilition (Group B) showed superior performance compared with no structured exercise (Group C). The overall hierarchy of functional benefit was A > B > C. Improvements in quality of life (KCCQ score) across all groups, including Group C, suggesting a subjective component. Adherence rates were moderate (72% in Group A, 83% in Group B, and 44% in Group C) and were assessed subjectively. Adverse events were minimal, non-serious, and not directly related to the interventions.
Conclusion
Individualized CR using AI was safe, feasible, and associated with improved functional outcomes in patients with HF when compared with conventional rehabilitation and with no structured exercise. The integration of AI into CR protocols represents a complementary and potentially scalable tool for clinical decision-making, particularly in settings with limited access to specialized resources, reinforcing its potential role in the future management of HF and in expanding the reach of healthcare services.