A Machine-Learning-Based Method to Detect Degradation of Motor Control Stability with Implications to Diagnosis of Presymptomatic Parkinson’s Disease: A Simulation Study

Abstract

Background and aim: Parkinson’s disease (PD), a neuro-degenerative disorder, is often detected by the onset of its motor symptoms such as rest tremor. Unfortunately, motor symptoms appear only when approximately 40–60% of the dopaminergic neurons in the substantia nigra are lost. In most cases, by the time PD is clinically diagnosed, the disease may already have started 4 to 6 years beforehand. There is therefore a need for developing a test for detecting PD before the onset of motor symptoms. This phase of PD is referred to as Presymptomatic PD (PPD). The motor symptoms of Parkinson’s Disease are manifestations of instability in the sensorimotor system that develops gradually due to the neurodegenerative process. In this paper, based on the above insight, we propose a new method that can potentially be used to detect the degradation of motor control stability, which can be employed for the detection of PPD. Methods: The proposed method tracks the tendency of a feedback control system to transition to an unstable state and uses a machine learning algorithm for its robust detection. This method is explored using a simple simulation example consisting of a simple pendulum with a proportional-integral-derivative (PID) controller as a conceptual representation for both healthy and PPD individuals with a noise variance of 0.01 and a noise variance of 0.1. The present study adopts a longitudinal design to evaluate the effectiveness of the proposed methodology. Specifically, the performance of the proposed approach, with specific choices of features, is compared to that of the Support Vector Machine (SVM) algorithm for machine learning under conditions of incremental delay-induced instability. This comparison is made with results obtained using the Longitudinal Support Vector Machine (LSVM) algorithm for machine learning, which is better suited for longitudinal studies. Results: The results of SVM with one choice of features are comparable with the results of LSVM for a noise variance of 0.01. These results are almost unaffected by a noise variance of 0.1. All of the methods showed a high sensitivity above 96% and specificity above 98% on a training data set. In addition, they perform very well with the validation synthetic data set with sensitivity above 95% and specificity above 98%. These results are robust to further increases in noise variance representing the large variances expected in patient populations. Conclusions: The proposed method is evaluated on a synthetic data set, and the machine learning results show a promise and potential for use for detecting PPD through an early diagnostic device. In addition, an example task with physiological measurement that can potentially be used as a clinical movement control test along with representative data from both healthy individuals and PD patients is also presented, demonstrating the feasibility of performing a longitudinal study to validate and test the robustness of the proposed method.