Accuracy of computer algorithms and the human eye in scoring actigraphy
Kathleen Boyne1, David D Sherry, Paul R Gallagher
1Divison of Pulmonology, The Children's Hospital of Philadelphia, Philadelphia, PA 19104, USA.
Sleep & Breathing = Schlaf & Atmung
|May 15, 2012
Summary
Human scoring of actigraphy, a sleep monitoring device, correlates better with polysomnography (PSG) than automated methods. Optimizing rest interval settings and using a low wake threshold improves accuracy for sleep analysis.
Area of Science:
- Sleep Science
- Biomedical Engineering
- Medical Devices
Background:
- Actigraphy is a non-invasive method for sleep monitoring.
- Polysomnography (PSG) is the gold standard for sleep studies.
- Comparing actigraphy scoring methods is crucial for accurate sleep assessment.
Purpose of the Study:
- To determine the optimal scoring method for actigraphy.
- To identify the best parameter settings for actigraphy analysis.
- To compare actigraphy performance against polysomnography.
Main Methods:
- Simultaneous polysomnography (PSG) and actigraphy data from 15 adolescent studies were analyzed.
- Actigraphy scoring by human visual inspection was compared to a commercial computerized algorithm.
- Various parameter settings, including rest period determination and wake thresholds, were evaluated.
Main Results:
- Human scoring of actigraphy showed a better correlation with PSG sleep parameters (r=0.640) than automated scoring (r=0.406).
- Setting rest intervals based on PSG yielded the highest correlation (r=0.694).
- Lower wake thresholds and 10-15 minute immobility times improved computerized scoring accuracy.
Conclusions:
- Visual actigraphy scoring demonstrates reasonable correlation with PSG.
- Accurate determination of the rest interval is critical for reliable actigraphy scoring.
- Human scoring outperforms automated methods when major rest periods are auto-set; manual adjustments and sleep diaries enhance accuracy.
