Multivariate classification of Brugada syndrome patients based on autonomic response to exercise testing

• Published in: PloS one 2018-05, Vol.13 (5), p.e0197367-e0197367
• Main Authors: Calvo, Mireia, Romero, Daniel, Le Rolle, Virginie, Béhar, Nathalie, Gomis, Pedro, Mabo, Philippe, Hernández, Alfredo I
• Format: Article
• Language: English
• Subjects: Adult; Aged; Analysis; Anàlisi multivariable; Artificial intelligence; Autonomic nervous system; Autonomic Nervous System - physiology; autonomic response; Bioengineering; Biology and Life Sciences; Biomedical engineering; Brugada syndrome; Brugada Syndrome - classification; Brugada Syndrome - physiopathology; Cardiac arrhythmia; Cardiology; Cardiology and cardiovascular system; Care and treatment; Case studies; Ciències de la salut; Classification; Computer and Information Sciences; Cor; Discriminant Analysis; Diseases; Electrocardiography; Engineering; Engineering Sciences; Exercise; Exercise Test; Family medical history; Feature extraction; Female; Funding; Genetic disorders; Heart; Heart rate; Heart Rate - physiology; HRV; Human health and pathology; Humans; International conferences; Life Sciences; Malalties; Male; Markers; Mathematical models; Medical tests; Medicina; Medicine and Health Sciences; Methods; Middle Aged; Modulation; Multivariate analysis; Neural networks; Parasympathetic nervous system; Patients; Physical Sciences; Physical training; Physiological aspects; Physiology; Populations; Recovery; Research and Analysis Methods; Risk factors; Signal and Image processing; Signal processing; Sleep; Software; Supervision; Ventricle; Young Adult; Àrees temàtiques de la UPC
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0197367

Ventricular arrhythmias in Brugada syndrome (BS) typically occur at rest and especially during sleep, suggesting that changes in the autonomic modulation may play an important role in arrhythmogenesis. The autonomic response to exercise and subsequent recovery was evaluated on 105 patients diagnosed with BS (twenty-four were symptomatic), by means of a time-frequency heart rate variability (HRV) analysis, so as to propose a novel predictive model capable of distinguishing symptomatic and asymptomatic BS populations. During incremental exercise, symptomatic patients showed higher HFnu values, probably related to an increased parasympathetic modulation, with respect to asymptomatic subjects. In addition, those extracted HRV features best distinguishing between populations were selected using a two-step feature selection approach, so as to build a linear discriminant analysis (LDA) classifier. The final features subset included one third of the total amount of extracted autonomic markers, mostly acquired during incremental exercise and active recovery, thus evidencing the relevance of these test segments in BS patients classification. The derived predictive model showed an improved performance with respect to previous works in the field (AUC = 0.92 ± 0.01; Se = 0.91 ± 0.06; Sp = 0.90 ± 0.05). Therefore, based on these findings, some of the analyzed HRV markers and the proposed model could be useful for risk stratification in Brugada syndrome.