Objective: To investigate whether markerless computer vision can objectively quantify axial symptoms in Parkinson’s disease (PD), focusing on posture and instability.

Background: Axial symptoms such as postural abnormalities and balance impairment are major contributors to disability and fall risk in PD. Clinical assessment mainly relies on semi quantitative rating scales that may lack objectivity. Computer vision and human pose estimation enable markerless motion analysis from low cost cameras, offering opportunities for scalable and remote monitoring.

Method: Forty PD patients (Hoehn and Yahr 2.5-4) were recruited across two Movement Disorders centers. Participants performed two 60 s quiet standing tasks (eyes open and eyes closed) recorded using a synchronized dual camera RGB-D setup. Skeletal models were extracted using Microsoft Azure Kinect (RGB-D reference) and Google MediaPipe Pose (RGB-only). Postural alignment parameters and center of mass sway metrics were computed from 3D trajectories and correlated (Spearman ρ) with clinical scales including MDS-UPDRS items and balance assessments.

Results: Lower complexity RGB-only models showed moderate to strong agreement with the RGB-D reference. Shoulder and hip horizontal alignment correlated with the reference system (ρ=0.57-0.73, p<0.001), while vertical trunk alignment showed strong consistency (ρ=0.79-0.87, p<0.001). Sagittal trunk and head angles from the lateral view also correlated with reference measurements (ρ>=0.72). Postural parameters were associated with axial impairment severity: frontal trunk alignment and sagittal trunk flexion correlated with MDS-UPDRS item 3.13 (ρ=0.35-0.55). Instability analysis based on body center trajectories showed moderate associations with clinical instability classification. Antero posterior sway parameters showed the strongest relationships, including total sway displacement (ρ≈0.46, p≈0.002) and mean sway velocity (ρ≈0.46–0.49, p<0.01), particularly during the eyes closed condition. Objective postural parameters also discriminated between clinical posture severity groups.

Conclusion: Markerless pose estimation enables objective quantification of axial symptoms in PD using simple standing tasks and supports camera based approaches for scalable quantitative assessment and remote monitoring.

References: This study was supported by “Objective monitoring of axial symptoms in Parkinson’s disease: quantitative assessment in daily life based on the use of wearables, video sensing and artificial intelligence (OMNIA-PARK)” project, funded by European Union – Next Generation EU within the PRIN 2022 PNRR program (D.D.1409 del 14/09/2022 Ministero dell’Università e della Ricerca). This abstract reflects only the authors’ views and opinions, and the Ministry cannot be considered responsible for them.

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/computer-vision-for-quantitative-assessment-of-posture-and-instability-in-parkinsons-disease/