A robust Parkinson's disease detection model based on time-varying synaptic efficacy function in spiking neural network

Abstract

Parkinson's disease (PD) is a neurodegenerative disease affecting millions of people around the world. Conventional PD detection algorithms are generally based on first and second-generation artificial neural network (ANN) models which consume high energy and have complex architecture. Considering these limitations, a time-varying synaptic efficacy function based leaky-integrate and fire neuron model, called SEFRON is used for the detection of PD. SEFRON explores the advantages of Spiking Neural Network (SNN) which is suitable for neuromorphic devices. To evaluate the performance of SEFRON, 2 publicly available standard datasets, namely (1) UCI: Oxford Parkinson's Disease Detection Dataset and (2) UCI: Parkinson Dataset with replicated acoustic features are used. The performance is compared with other well-known neural network models: Multilayer Perceptron Neural Network (MLP-NN), Radial Basis Function Neural Network (RBF-NN), Recurrent Neural Network (RNN) and Long short-term memory (LSTM). The experimental results demonstrate that the SEFRON classifier achieves a maximum accuracy of 100% and an average accuracy of 99.49% on dataset 1. For dataset 2, it attains a peak accuracy of 94% and an average accuracy of 91.94%, outperforming the other classifiers in both cases. From the performance, it is proved that the presented model can help to develop a robust automated PD detection device that can assist the physicians to diagnose the disease at its early stage.

Keywords: Parkinson’s disease; SEFRON; Spike timing dependent plasticity (STDP); Spiking neural network; Time-varying synaptic efficacy function.

Fig. 1

Correlation Matrix of the features present in dataset#1

Fig. 2

Correlation Matrix of the features present in dataset#2

Fig. 3

General block diagram for the detection of PD using neural network models

Fig. 4

Architecture of SEFRON model with P number of presynaptic neuron and a single postsynaptic neuron

Fig. 5

Impact of number of receptive field neurons ($Q$) on the Gmean [Dataset#1] of SEFRON

Fig. 6

Impact of overlap constant ($\beta$) on the Gmean [Dataset#1] of SEFRON

Fig. 7

Impact of learning rate ($\mu$) on the Gmean [Dataset#1] of SEFRON

Fig. 8

Impact of sigma (σ) on the Gmean [Dataset#1] of SEFRON

Fig. 9

Impact of STDP learning window (${\tau }_{+}$) on the Gmean [Dataset#1] of SEFRON

Fig. 10

Impact of time constant ($\tau$) on the Gmean [Dataset#1] of SEFRON

Fig. 11

Performance comparison of MLP, RBF, RNN, LSTM and SEFRON using K-fold cross validation with different value of ‘K’ for dataset#1

Fig. 12

Performance comparison of MLP, RBF, RNN, LSTM and SEFRON with different percentage split of training and testing sets for dataset#1

Fig. 13

Comparison of box plots of (a) accuracy, (b) sensitivity and (c) specificity for MLP, RBF, RNN, LSTM and SEFRON using 10-fold cross validation for dataset #1