A deep learning based model using RNN-LSTM for the Detection of Schizophrenia from EEG data

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Schizophrenia diagnosis can be challenging, impacting patient quality of life.
• Electroencephalogram (EEG) signals reflect brain network connectivity, potentially revealing anomalies associated with schizophrenia.
• Deep learning offers automated feature extraction and classification capabilities for complex biological data.

Purpose of the Study:

• To develop and evaluate a deep learning model for diagnosing schizophrenia using EEG data.
• To assess the model's performance against traditional machine learning classifiers and existing deep learning approaches.

Main Methods:

• A Recurrent Neural Network-Long Short-Term Memory (RNN-LSTM) deep learning model was designed with three dense layers.
• EEG data from 45 schizophrenic patients and 39 healthy subjects were analyzed.
• Dimensionality reduction techniques were applied to optimize feature sets for classification.

Main Results:

• The proposed RNN-LSTM model achieved 98% accuracy with a complete feature set and 93.67% with a reduced feature set.
• The model demonstrated superior performance compared to Random Forest, SVM, FURIA, and AdaBoost classifiers.
• Accuracy was comparable or better than existing Convolutional Neural Network (CNN) or RNN models using the same dataset.

Conclusions:

• Deep learning, specifically RNN-LSTM, is a highly effective tool for diagnosing schizophrenia from EEG signals.
• The proposed model offers a robust and accurate method for early schizophrenia detection.
• This approach holds promise for improving diagnostic accuracy and patient outcomes.