A method for automatic, objective and continuous scoring of bradykinesia

The assessment of bradykinesia is a key element in the diagnosis of Parkinson's disease. It is typically performed using the Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS). However, despite its importance, the bradykinesia-related items of this scale show very low inter-rater agreement. Therefore, in this study a method for automatic, objective and continuous scoring of three of the bradykinesia-related items of the MDS-UPDRS is proposed. Four clinicians scored these items for 25 patients diagnosed with Parkinson's disease, within a range of 0-4. Orientation sensors were used to record movement during performance of each item. From the recorded data a set of features was derived to represent the movement characteristics that evaluators assess for scoring bradykinesia according to the MDS-UPDRS. These features and the averaged scores of the evaluators were used to create a model for the score on each item using backward linear regression. The estimated generalization errors indicate that the continuous objective scale can obtain an automatic score with an average error of 0.50 compared to the evaluators' averaged scores.