• Citer

A new non-contact method to detect heart rate and explore aortic valve disease semiology

par Abdessater, Elza ;Creismeas, Maël;Balali, Paniz ;Rabineau, Jérémy ;Tordeur, Cyril ;Faoro, Vitalie ;Van De Borne, Philippe ;Hossein, Amin
Référence 44th Annual Congress of the Belgian Society of Cardiology (6 and 7 February 2025: Brussels)
Publication A Paraître, 2025-02-06

Poster de conférence

• Détails
• Contenu

Résumé :

Introduction : The rise in smartphone technology highlights its potential as a cost-effective medical tool. Some studies suggest that smartphone cameras enable heart rate (HR) measurements through facial skin color variations using remote photoplethysmography (rPPG). These color changes evoke the Lighthouse sign observed in severe aortic regurgitation (AR), indicating that rPPG may become a tool for augmented semiology.Purpose : This study assesses the accuracy of HR obtained from smartphone videos.Method : In a monocentric prospective study, patients with severe aortic valve disease (AVD) and healthy subjects of various phototypes were recruited, excluding those with arrhythmia. Two 30-second facial videos were recorded for each subject using a smartphone (4K resolution, 60 frames per second), with subjects seated 30 cm away from the camera. A temporal signal was extracted from each video and analyzed to detect HR through an improved rPPG method. Simultaneously, a single-lead ECG recorded at 500Hz served as reference to compare HR and assess root-mean-square error (RMSE), correlation coefficient (CC), coefficient of variation (CV), and intraclass correlation coefficient (ICC).Results : A total of 33 subjects were included: 26 healthy subjects and 7 subjects with AVD (4 severe aortic stenosis, 1 severe AR, 2 severe mixed AVD). In healthy subjects, HR obtained by rPPG versus ECG showed an RMSE of 0.78 bpm, a CC of 0.99, and an ICC of 0.99 (95% CI 0.99-1.00). For AVD patients, an RMSE of 1.05 bpm, a CC of 0.99, and an ICC of 0.99 (95% CI 0.99-1.00) were observed.Conclusion : HR can be reliably determined using an improved rPPG method, in both healthy subjects and AVD patients. The temporal signal extracted from the video may hold potential for future AVD detection methods.