Mathematically Modeling the GPe/STN Neuronal Cluster to Account for Parkinsonian Tremor and Developing a Novel Method to Accurately Diagnose Parkinson's Disease Using Speech Measurements and an Artificial Neural Network

TL;DR

This paper introduces a novel speech-based neural network diagnostic tool for Parkinson's disease and a mathematical model of neural clusters to explain tremors, advancing early detection and understanding of PD.

Contribution

It develops a speech analysis neural network for PD diagnosis with 96.55% accuracy and constructs a mathematical neural cluster model explaining Parkinsonian tremors.

Findings

Neural network correctly classifies 96.55% of test speech data.

Mathematical model predicts biological neuron firing patterns in PD.

Model accounts for neuron bursting, neurotransmitter release, and synaptic connections.

Abstract

Parkinsons disease (PD) is a debilitating motor system disorder characterized by progressive loss of movement, tremors, and speech slurring. PD is due to the loss of dopamine-producing brain cells, and symptoms only worsen over time, making early detection and diagnosis of the disease key to effective management and treatment. There is currently no standardized method of diagnosis available, and instead a combination of the medical history of a patient and physician judgment is used. In this research, a novel method of accurately diagnosing PD using an artificial neural network (ANN) and speech measurements was developed. Using this technology, a patient need only speak into a computer microphone. Speech data is then analyzed using Praat and inputted into the ANN to obtain the diagnosis. The ANN, built using MATLAB and trained and tested with actual patient data, was able to correctly classify 96.55% of test data. A mathematical model of the GPe/STN neural cluster was then constructed to account for Parkinsonian tremor. This was done using a two-cell model that coupled a GPe neuron to a STN neuron and consisted of a system of twelve ordinary differential equations. The model was structured to account for neuron bursting behavior, neurotransmitter release, and synaptic connections between neurons. After comparison to the biological behavior of the cluster and neuron firing patterns in PD patients, the model was determined to be predictive of known biological results. This work presents a significant step forward in PD research and could be successfully implemented into clinical practice.