Study On The Detection And Recognition Of Tics In Tourette Syndrome Based On Deep Learning

Tourette syndrome(TS)is a neurological disorder characterized by repetitive,stereotyped,involuntary movements and vocalizations called tics,which is prone to depression,anxiety,sleep problems and other complications,and has a serious impact on the normal life and social interaction of patients.TS is a diagnosis that doctors make after history taking,clinical examination for a long time.The evaluation of tic symptom severity requires doctors to interview patients and score tics according to the TS scale,which takes a lot of time and requires a high degree of coordination between doctors and patients.At the same time,patients may hide the tics in the interview,which is not conducive to the diagnosis and evaluation.The research explored the feasibility of deep learning in the tic detection and recognition during the TS.The dataset contains 72 patients diagnosed with TS from the pediatrics clinic of the Second Affiliated Hospital of Zhejiang University School of Medicine.For each individual,there was a video shot from the front.A ResNet model based on multiple instance learning(MIL)to detect the abnormal actions from the video and a three-dimensional convolutional network(3D-CNN)for tic actions recognition were built.The results showed that the first model achieved the accuracy of about 95.48%in tic detection,and the accuracy of classification model based on 3D-CNN reached 89.5%.Besides,the accuracy of testing on the unlabeled dataset reached 77.03%and the recall reached 83.81%,indicating that the model has good generalization ability.To sum up,the results of the research proved the feasibility of deep learning in tic actions detection and recognition,which could effectively help doctors to labeling actions in videos in an automatic way,and provide reference for the auxiliary diagnosis and severity evaluation of TS.