Abstract:
Clinical criteria—bradykinesia, rest tremor, rigidity—diagnose Parkinson’s disease (PD). Clinical rating scales for PD severity are subject to inter-rater variability. This paper proposes a deep learning-based automatic PD diagnosis method using videos for clinical use. We demonstrate the efficacy of a 3D CNN for PD severity classification. Transfer learning from non-medical datasets can help PD severity classification due to the lack of clinical data. To bridge the domain discrepancy between medical and non-medical datasets, we designed a Temporal Self-Attention (TSA) mechanism to focus the network on subtle temporal visual cues, such as tremor frequency. Seven MDS-UPDRS part III tasks reveal bradykinesia and postural tremors. Task-assembling can predict patient-level PD severity using multi-domain learning. Our PD video dataset empirically demonstrates TSA and task-assembling method effectiveness. Our best MCCs are 0.55 on binary task-level classification and 0.39 on three-class patient-level classification.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
