Photoplethysmography in recent-onset atrial fibrillation: automatic detection of rhythm change and burden

Figures

• 

  Figure 1. Data collection flowchart. The final study population included 90 atrial fibrillation and 90 sinus rhythm episodes from 90 patients. Initially, the study protocol was tested with 10 patients. After obtaining written consent, 24 patients were excluded due to missing acceleration data (2) or photoplethysmography data (8) as a result of the initial protocol testing or because they presented with atrial flutter.14 Additionally, six patients were excluded because the cardioversion was cancelled (one) or the rhythm converted spontaneously to sinus rhythm just after recruitment, resulting in less than 5 min of photoplethysmography data being recorded during atrial fibrillation (five).

• 

  Figure 2. Distribution of CHA2DS2-VA scores.

• 

  Figure 3. Distribution of atrial fibrillation and sinus rhythm recordings.

• 

  Figure 4. Analysable data according to the detector window. The values are presented as percentages and sorted as patient index based on the proportion of analysable data of each patient as determined by a quality algorithm.

• 

  Figure 5. (A) Time-based atrial fibrillation detection: sensitivity, specificity, and (B) delay in detecting rhythm change after cardioversion according to the detector window. Patient Index: number of patients. (A) The sensitivity and specificity values are presented as percentages and sorted by sensitivity. (B) Sinus rhythm detected indicates the delay after cardioversion when the rhythm change was detected by the automated method.

• 

  Figure 6. Bland–Altman plots comparing photoplethysmography and electrocardiogram atrial fibrillation burdens according to detector window length. Values are expressed as duration in minutes. The x-axis represents the mean of the photoplethysmography and electrocardiogram values, and the y-axis represents the difference between the electrocardiogram and photoplethysmography values.