Motor coordination is the combination of body movements created with the kinematic (spatial direction) and kinetic (force) parameters that result in intended actions Trigger finger is a common symptom in hand that affects the motor coordination and results in hand functional impairment With hand functional impairment our movements cannot reach a certain level of motion quality to carry out normal activities Previous studies used cylindrical grasp devices to investigate finger force coordination of the hand during precision grasping However data acquired is usually dynamic multivariate and high dimensional While both machine learning and deep learning are powerful tools to analyze these complex data they also have differences Machine learning uses pre-defined features that have clinical meaning verified in previous studies to train the model Deep learning use kernels to extract feature automatically during the training process therefore these features are still unknown and should be discussed This study proposes a machine learning and deep learning approach to investigating finger forces coordination during grasping task and drinking task 44 healthy subjects (39 5 years ± 7 6) and 54 trigger finger patients (57 6 years ± 8 0) participated in this study We built and analyzed 2 supervised classification models (Random Forest and 1D-CNN) Random Forest gives feature importance from all extracted features As for 1D-CNN the Grad-CAM technique gives information on which phase is the most important for the model to make the classification This study found out that Random Forest and 1D-CNN model can classify between two groups of subjects with an average accuracy of ~77% Random Forest results suggest that the duration of trials is an important factor while not controlled in the experiment setting 1D-CNN results suggest that the holding phase is the most important phase for the model to make the classification
Date of Award | 2020 |
---|
Original language | English |
---|
Supervisor | Fong-chin Su (Supervisor) |
---|
Multi-digit Coordination in patient with Trigger Digit during Natural Precision Grasping using Machine Learning and Deep Learning
堅, 陳. (Author). 2020
Student thesis: Doctoral Thesis