The automatic detection method for hand tremors in Parkinson’s disease based on video analysis

Hand tremor is a primary characteristic of Parkinson’s disease. Accurately assessing the severity of hand tremors is crucial for the treatment and care of Parkinson’s patients. This paper proposes an automatic detection method based on video analysis, accurately detecting hand tremors from videos by applying image processing algorithms and motion analysis techniques. The method integrates frequency features from pose estimation and image pixel change features from three-frame differences. Machine learning classifiers are applied to automatically detect tremors and non-tremors. We evaluated our framework in a large dataset of 98 Parkinson’s patients. Our method achieved accuracy of 98.3% and 95.3% for rest tremor and postural tremor, respectively. Regarding the classification of hand tremor severity, our method achieved a precision of 98.4% and 99.1% for rest tremor and postural tremor. Experimental results indicate that our proposed method can accurately detect hand tremors in Parkinson’s patients and is consistent with clinical expert assessments. Our research demonstrates the effectiveness of computerassisted assessment based on video analysis for tremors in Parkinson’s disease.