Recurrence analysis of cardiac arrhythmias: A systematic review of methods, substrate characterisation and clinical translation

Cardiac arrhythmias are a major cause of morbidity and mortality, and catheter ablation depends on accurate identification of the underlying substrate. Cardiac signals are nonlinear and non-stationary, which limits Fourier-based analysis. Recurrence quantification analysis (RQA) and related recurrence methods make fewer assumptions, but their use in arrhythmia research has not been systematically synthesised. This review examines what recurrence analysis has demonstrated, how consistently its methods have been applied, and what remains before it can support clinical substrate characterisation. Of 309 records retrieved from four databases (207 after deduplication), 48 studies met the eligibility criteria and citation searching added one, giving 49. Each study was examined in full text for its signals, cohorts, embedding parameters, threshold selection, metrics and results, and assigned to one of six methodological categories. The three studies that tested recurrence against surrogate or shuffled data each found structure in fibrillatory signals that matched stochastic surrogates did not reproduce. Recurrence metrics and dominant frequency were weakly correlated and diverged in substrate characterisation, so they are not interchangeable. Threshold selection was the most consequential uncontrolled variable: five strategies were in use, ten of the 35 studies that apply a threshold reported no selection rule, and only three examined how their results depended on it. Determinism and laminarity, although the most frequently reported metrics, were not consistently the strongest discriminators. Clinical readiness is not established: only one study used recurrence mapping prospectively to guide ablation, in ten patients without randomisation. A shared threshold-selection standard and prospective multicentre validation are needed.