Machine Learning Enabled Time Series Analysis in Medicine

Pattern Recognition in Heart Rate Variability Using Fitness Trackers in Cardiovascular Disease

The goal of this observational cohort study is to investigate the potential of fitness trackers in combination with machine learning algorithms to identify cardiovascular disease specific patterns.

Two hundred participants will be enrolled:

1. 50 with heart failure
2. 50 with atrial fibrillation
3. 100 (healthy) individuals without the former two conditions

All participants are given a Fitbit device and monitored for three months. Researchers will compare differences in heart rate variability patterns between the groups and devise a machine learning algorithm to detect these patterns automatically.

Inclusion Criteria:

• systolic heart failure (LVEF < 35%)
• Atrial fibrillation without heart failure
• Individuals without cardiovascular disease

Exclusion Criteria:

• > 85 years old
• Recent pulmonary venous antrum isolation procedure (<1 year)
• (end stage) kidney failure
• (end stage) liver failure
• Study participants with known systemic active inflammatory disease
• Study participants with impaired mental state
• Inability to use a fitness tracker or mobile phone
• Impaired cognition and inability to understand the study protocol