Background: Programming deep brain stimulation (DBS) of the subthalamic nucleus for optimal symptom control in Parkinson's Disease (PD) requires time and trained personnel. Novel implantable neurostimulators allow local field potentials (LFP) recording, which could be used to identify the optimal (chronic) stimulation contact. However, literature is inconclusive on which LFP features and prediction techniques are most effective. Objective: To evaluate the performance of different LFP-based physiomarkers for predicting the optimal (chronic) stimulation contacts. Methods: A literature search was conducted across nine databases, resulting in 418 individual papers. Two independent reviewers screened the articles based on title, abstract, and full text. The quality of included studies was assessed using a modified Joanna Briggs Institute Critical Appraisal Checklist for Case Series. Results were categorised in four classes based on the predictive performance with respect to the a priori chance. Results: Twenty-five studies were included. Single-feature beta-band predictions demonstrated positive performance scores in 94 % of the outcomes. Predictions based on single non-beta-frequency features yielded positive scores in only 25 % of the outcomes, with positive results mainly for high frequency oscillations. Multi-feature predictions (e.g. machine learning) achieved accuracy scores within the two highest performance classes more often than single beta-based predictions (100 % versus 39 %). Conclusion: Predicting the optimal stimulation contact based on LFP recordings is feasible and can improve DBS programming efficiency in PD. Single beta-band predictions show more promising results than non-beta-frequency features alone, but are outperformed by multi-feature predictions. Future research should further explore multi-feature predictions for optimal contact identification.

Parkinson's disease (PD) diagnosis is based on clinical criteria, i.e., bradykinesia, rest tremor, rigidity, etc. Assessment of the severity of PD symptoms with clinical rating scales, however, is subject to inter-rater variability. In this paper, we propose a deep learning based automatic PD diagnosis method using videos to assist the diagnosis in clinical practices. We deploy a 3D Convolutional Neural Network (CNN) as the baseline approach for the PD severity classification and show the effectiveness. Due to the lack of data in clinical field, we explore the possibility of transfer learning from non-medical dataset and show that PD severity classification can benefit from it. To bridge the domain discrepancy between medical and non-medical datasets, we let the network focus more on the subtle temporal visual cues, i.e., the frequency of tremors, by designing a Temporal Self-Attention (TSA) mechanism. Seven tasks from the Movement Disorders Society - Unified PD rating scale (MDS-UPDRS) part III are investigated, which reveal the symptoms of bradykinesia and postural tremors. Furthermore, we propose a multi-domain learning method to predict the patient-level PD severity through task-assembling. We show the effectiveness of TSA and task-assembling method on our PD video dataset empirically. We achieve the best MCC of 0.55 on binary task-level and 0.39 on three-class patient-level classification.