Automatic two-dimensional & three-dimensional video analysis with deep learning for movement disorders: A systematic...

Area of Science:

• Neurology
• Computer Science
• Biomedical Engineering

Background:

• Traditional movement disorder diagnosis relies on subjective assessments, often missing early or overlapping symptoms.
• Computer vision and deep learning offer objective, quantitative analysis of motor symptoms.
• Video analysis provides a practical, low-cost solution for movement disorder assessment.

Purpose of the Study:

• To systematically review advancements in deep learning-based video analysis for movement disorders.
• To consolidate knowledge on objective video analysis for conditions like Parkinson's disease, ataxia, and Tourette syndrome.
• To identify key methodologies, datasets, and challenges in the field.

Main Methods:

• Systematic literature review of studies published until September 2023.
• Searched major scientific databases (Web of Science, PubMed, Scopus, Embase).
• Analyzed 68 relevant studies focusing on objectives, datasets, modalities, and deep learning techniques.

Main Results:

• Identified diverse applications including Parkinson's disease symptom quantification, ataxia assessment, and tic detection.
• Examined various video modalities (2D/3D) and deep learning architectures used.
• Cataloged datasets and methodologies for objective movement disorder analysis.

Conclusions:

• Deep learning-powered video analysis presents significant advancements for objective movement disorder diagnosis.
• Opportunities exist in improving datasets, interpretability, and enabling remote monitoring.
• Further research is needed to address current challenges and enhance clinical utility.