Quantification of Parkinsonian Kinematic Patterns in Body-Segment Regions During Locomotion

Diagnosis and treatment of Parkinson’s Disease (PD) are typically supported by a kinematic gait analysis. Nonetheless, the main drawbacks classical analysis, based on reduced set markers, loss small dynamical changes, invasive methodology, sparse representation from few points, restricting disease This work aims to perform robust regional characterization, which may result in potential digital biomarker complement personalized monitoring PD. introduces markerless computational framework full body-segment characterization related with PD motor alterations. Firstly, dense motion trajectories computed represent locomotion. Such grouped using deep learning body segmentation, that partitions human silhouette into regions corresponding head, trunk limbs. Each resultant region is described dartboard-like histograms along trajectories. The proposed approach was validated different pretrained classification models. method evaluated 11 control subjects patients, achieving an average accuracy $$99.62\%$$ for lower-limbs head regions. proved be effective classify Parkinsonian patterns w.r.t gaits. A major contribution strategy capability recover segments, particularly, regions, turned out decisive biomarker.