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

Area of Science:

• Neurology
• Computer Science
• Biomedical Engineering

Background:

• Parkinson's disease (PD) affects millions globally, with current detection relying on energy-intensive, complex artificial neural network (ANN) models.
• Limitations of conventional ANNs include high energy consumption and intricate architectures, hindering efficient PD diagnosis.

Purpose of the Study:

• To introduce and evaluate SEFRON, a novel Spiking Neural Network (SNN) model for Parkinson's disease detection.
• To compare SEFRON's performance against established neural network models for PD detection.

Main Methods:

• Utilized a leaky-integrate and fire neuron model with time-varying synaptic efficacy (SEFRON) for PD detection.
• Evaluated SEFRON on two standard datasets: UCI Parkinson's Disease Detection Dataset and UCI Parkinson Dataset with replicated acoustic features.
• Compared SEFRON against Multilayer Perceptron Neural Network (MLP-NN), Radial Basis Function Neural Network (RBF-NN), Recurrent Neural Network (RNN), and Long short-term memory (LSTM).

Main Results:

• SEFRON achieved 100% maximum and 99.49% average accuracy on the first dataset.
• SEFRON attained 94% peak and 91.94% average accuracy on the second dataset.
• SEFRON outperformed all compared neural network models in accuracy on both datasets.

Conclusions:

• SEFRON demonstrates superior performance in Parkinson's disease detection compared to conventional neural networks.
• The SEFRON model shows potential for developing robust, automated PD detection devices for early diagnosis assistance.
• This SNN-based approach is suitable for neuromorphic devices, offering energy efficiency and simpler architectures.