Wearable Sensor And Classification Algorithms-based Automatic Objective Scoring Of Motor Symptoms In Parkinson

Objective:Based on the wearable device and support vector machine(SVM)to automatically and objectively evaluate the score results of the finger tapping test in UPDRS bradykinesia,it is expected to obtain an objective bradykinetic score to eliminate clinical inconsistent assessment results.Methods:In this study,a total of 79 items of finger tapping test data of hands were collected from 40 subjects.Data were collected using two portable motion sensors mounted on subjects’ thumbs and index fingers.Subjects were asked to perform a 30-second finger tapping test,and at the same time,finger tapping items were clinically graded by the unified UPDRS.Randomly selected article 60 of them as the training data,19 as prediction data.Accuracy was selected as the default index to evaluate the best model,and recall rate,kappa coefficient,f1-score and mean square error(MSE)were calculated.Support vector machine(SVM)was used to construct the model,and the influence of different dimensions of the two sensors and the time of data on the model was compared..Results:A total of 79 evaluations of both hands were recorded,of which PD FT evaluation was 0 points for 35 items,1 point for 35 items,2 points for 20 items and 3 points for 10 items.The results show that in the test set,the correct rate is 73.7%,the 0s-10 s correct rate is 47.4%,the 10s-20 s correct rate is 84.2%,and the 20s-30 s correct rate is 63.2%.For the second piece of data,compare the effects of different dimensions of the two sensors on the model.Since each sensor has three dimensions,X,Y,and Z,this experiment only compares whether the different dimensions of a single sensor have an impact on the model.The results show that the accuracy of the Y2 + Z2 channel is 68.4%.Therefore,after extracting the data of all dimensions of the two sensors on the index finger and thumb as the feature input,the SVM model has a high accuracy.Conclusion:In our study,we demonstrated the objective evaluation of FT based on the sensor and SVM classification method,and obtained the consistency evaluation.The result showed that the accuracy of the evaluation was 73.7% in the whole period,and that of the second period(20-20 seconds)was 84.2%.The results show that the UPDRS score finger tapping test can be effectively graded by using the motion sensor and SVM classification method,using two sensors and 10-20 seconds acquisition time.