A simple model to detect atrial fibrillation via visual imaging.

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Atrial fibrillation (AF) detection is a significant clinical challenge.
• Current methods may require specialized equipment or direct patient contact.
• Developing non-invasive and accessible AF screening tools is crucial.

Purpose of the Study:

• To propose and validate a novel model for identifying atrial fibrillation (AF) using remote facial video recordings.
• To assess the efficacy of photoplethysmographic imaging (PPGi) signals extracted from facial videos for AF detection.
• To differentiate AF from sinus rhythm (SR) and atrial flutter/frequent ectopic beats (ARR).

Main Methods:

• Analysis of photoplethysmographic imaging (PPGi) signals derived from facial videos of 68 patients (30 SR, 25 AF, 13 ARR).
• Computation of 26 distinct physiological indexes, including Mean of inter-systolic interval series (M), Local Maxima Similarity (LMS), and Pulse Harmonic Strength (PHS).
• Development and validation of a classification model using a three-feature subset (M, Similarity1, LMS) on training, validation, and test datasets.

Main Results:

• Significant differences in M, LMS, and PHS indexes were observed among SR, AF, and ARR groups.
• Variability and irregularity parameters were lowest in SR, highest in AF, and intermediate in ARR.
• The validated model achieved high accuracy rates: 0.876 for SR, 0.870 for AF, and 0.863 for ARR on unseen test data.

Conclusions:

• Contactless, video-based monitoring using facial PPGi signals is a viable method for detecting atrial fibrillation (AF).
• The developed model effectively differentiates AF from sinus rhythm (SR) and other arrhythmias like atrial flutter or frequent ectopic beats (ARR).
• This technology offers a promising non-invasive approach for widespread AF screening and monitoring.