Fig. 6. Examples from the FastEval Parkinsonism system.

a The illustration depicts the outcome of a one-click analysis. Keypoint skeletons were generated using MediaPipe to annotate the provided video. In the upper middle position, the evolving frequency and intensity over time are showcased, effectively capturing the motor movement dynamics during recording. In the middle, the normalized distances between the index finger’s tip and the thumb tip are illustrated, with annotated detected peaks. The middle’s bottom plot exhibits the absolute frequency difference, serving as a potential indicator of interruptions or hesitations. This is due to the noticeable frequency change in case of motor movement interruptions. On the right panel, additional digital details are presented, encompassing the assessed hand, confidence level, estimated MDS-UPDRS item score, and hand parameters (evaluation indices). b The showcased radar plot serves as a clear example, vividly depicting the contrast in motor movement severity between the left and right hands. This distinction is achieved by employing four distinct hand parameters alongside a label estimated by a deep-learning model. Notably, both the hand parameters and the label undergo linear transformation to an 80–20 scale, using the median of our cohort dataset as the reference point. This visualization, displaying both hands simultaneously, proves instrumental for clinicians in conducting a rigorous quantitative evaluation of severity. Additionally, it holds the potential to facilitate the early diagnosis of atypical parkinsonism cases. c This example offers a representative illustration of time-dependent tracking on a finger-tapping hand parameter, frequency. Through this visualization, clinicians can effectively assess the progression between two clinical visits and potentially adjust medication dosages to enhance the efficacy of treatment plans.