# Beyond burden metrics: Wearable photoplethysmography-derived spatiotemporal progression of atrial fibrillation linked to clinical outcomes

## TL;DR

This study uses wearable sensors to track how atrial fibrillation progresses over time, linking these patterns to patient outcomes like symptoms and treatment needs.

## Contribution

A novel spatiotemporal model using wearable PPG data to dynamically quantify atrial fibrillation progression and risk.

## Key findings

- A 5-dimensional model showed strong agreement with Holter monitoring for AF episode duration (r = 0.97).
- A composite AF burden score ≥0.59 identified high-risk patients for symptoms or heart rate issues.
- AF progression varied by treatment type and symptom severity, with ablation-treated patients showing shorter episode durations.

## Abstract

Traditional atrial fibrillation (AF) classification lacks dynamic quantification. Current AF burden assessment is constrained by intermittent monitoring and simplistic metrics.

This study aimed to establish a continuous, multidimensional AF progression model using wearable photoplethysmography (PPG) for real-world, dynamic burden quantification.

In this prospective cohort, 110 patients with paroxysmal AF underwent synchronized PPG (Huawei Watch GT3) and 24-hour Holter monitoring. We developed a multiscale fusion AF algorithm and a 5-dimensional spatiotemporal progression model quantifying episode frequency, duration, clustering, circadian rhythm, and tachycardia burden.

The fusion algorithm achieved an accuracy of 0.97. The 5-dimensional model showed strong concordance with Holter monitoring, with near-perfect correlation for episode duration (r = 0.97) and high interchangeability (intraclass correlation coefficient >0.75). It demonstrated excellent diagnostic performance for burden trajectories (area under the curve 0.98). A composite AF burden score of ≥0.59 identified patients at high risk of AF-related symptoms or heart rate issues. Clinically, AF burden increased with worsening European Heart Rhythm Association symptoms (P = .04), varied by risk profile (P = .03), and differed between ablation- and drug-treated patients (episode duration P < .001).

Continuous PPG-based spatiotemporal modeling robustly quantified dynamic AF progression, enabled precise phenotyping, and may support early intervention in high-risk patients.