Deep Learning-Based Image Automatic Assessment and Nursing of Upper Limb Motor Function in Stroke Patients

This paper mainly introduces the relevant contents of automatic assessment of upper limb mobility after stroke, including the relevant knowledge of clinical assessment of upper limb mobility, Kinect sensor to realize spatial location tracking of upper limb bone points, and GCRNN model construction process. Through the detailed analysis of all FMA evaluation items, a unique experimental data acquisition environment and evaluation tasks were set up, and the results of FMA prediction using bone point data of each evaluation task were obtained. Through different number and combination of tasks, the best coefficient of determination was achieved when task 1, task 2, and task 5 were simultaneously used as input for FMA prediction. At the same time, in order to verify the superior performance of the proposed method, a comparative experiment was set with LSTM, CNN, and other deep learning algorithms widely used. Conclusion. GCRNN was able to extract the motion features of the upper limb during the process of movement from the two dimensions of space and time and finally reached the best prediction performance with a coefficient of determination of 0.89.