A Pilot Study: Automatically Monitoring the Physical Dysfunction of Idiopathic REM Sleep Behaviour Disorder with Wearables and Deep Learning.

Idiopathic Rapid eye movement (REM) sleep behavior disorder (iRBD) is considered a marker of Parkinson's disease (PD). The diagnosis and follow-up approach is video polysomnography (vPSG), which is too expensive to be accessible to the general population, and its precision could be affected by the first-night effect. Therefore, this study explores an economic approach to continuously monitoring iRBD at home. In accordance with the findings of previous studies, which suggest that iRBD exhibits symptoms of motor dysfunction similar to PD, this study proposes automatically recognizing the physical examinations performed by patients with iRBD and statistically analyzing their performance. To achieve this, 11 RBD patients were recruited and required to undergo eight physical examinations using the Movement Disorder Society-sponsored unified Parkinson's disease rating scale (MDS-UPDRS) part III for activity monitoring. Their movement signals were collected by five accelerometers, attached separately to the wrists, ankles, and abdomen. A novel channel-attention integrated temporal convolutional network, STCN, was developed to recognize activities, achieving an F1 score of 0.939 ± 0.106. During the iRBD performance analysis, a public dataset, PD-BIOSTAMPRC21, was also included, including 17 patients with PD and 14 healthy controls performing the same activities. Seven types of features were extracted from the accelerometer signals from every activity. The results suggested that among the three groups, there are significant differences in performing certain examinations, such as finger tapping(p = 0.046) and toe tapping (p = 0.049). This study verifies the potential of automatically monitoring patients with iRBD by tracking their performance in the tests of UPDRS III.Clinical relevance- This study verifies the potential of monitoring iRBD via wearables and the deep learning algorithm.