Researchers have developed a deep learning approach to estimate turning angles for Parkinson's disease (PD) severity assessment using single monocular camera videos. The method extracts 3D skeletons from videos, calculates hip and knee joint rotations, and achieves a turning calculation accuracy of 41.6% with a Mean Absolute Error of 34.7° on a dataset of free-living home videos. This work is the first to quantify PD patient turns in a home setting, aiming to provide more sensitive indicators of disease progression than traditional clinic-based assessments. AI
IMPACT Enables continuous, passive monitoring of Parkinson's disease progression in home environments, potentially leading to more personalized treatment and faster symptom detection.
RANK_REASON Academic paper detailing a new method for disease assessment. [lever_c_demoted from research: ic=1 ai=1.0]
- Fastpose
- Human3.6M
- Parkinson's disease
- Qiushuo Cheng
- Strided Transformer
- Turn-H3.6M
- Turn-REMAP
- Your Turn
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
