International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 09 | Sep 2024
p-ISSN: 2395-0072
www.irjet.net
DeprNet: A Deep Convolutional Neural Network for EEGBased Depression Detection 1 K.VENKATESWARLU, 2YELLAPU MARY IMMACULATA TWINKLE 1Department of Computer Science and Systems Engineering, Andhra University College of Engineering,Andhra
University, Visakhapatnam,Andhra Pradesh,India. 2Department of Computer Science and Systems Engineering, Andhra University College of Engineering,Andhra University, Visakhapatnam,Andhra Pradesh,India ---------------------------------------------------------------------***-------------------------------------------------------------------Abstract: Depression is a leading factor contributing to the global increase in suicide rates, emphasizing the need for its timely and accurate diagnosis and treatment. This paper presents DeprNet (a deep convolutional neural network), a novel deep learning framework based on convolutional neural networks (CNNs) for classifying EEG(electroencephalography) data of individuals with and without depression. The severity of depression is assessed using the Patient Health Questionnaire 9 (PHQ9) scale. DeprNet achieves remarkable performance, with an accuracy of 0.9937 and an AUC score of 0.999 for recordwise split data, and an accuracy of 0.914 with an AUC of 0.956 for subjectwise split data. These findings suggest that Convolutional Neural Networks (in short we call it as CNNs) can be prone to overfitting when the dataset includes limited subjects. However, DeprNet surpasses eight baseline models in performance, particularly highlighting the importance of EEG signals from the right and left electrodes in differentiating between depressed and nondepressed subjects. Keywords: Depression Detection, Deep Neural Networks, Convolutional Neural Network(in short we call it as CNN), Patient Health Questionnaire (PHQ9),EEG Signals.
1. INTRODUCTION Depression has emerged as a major mental health concern around the world, impacting millions of people and ranking among the top causes of disability. Its widespread influence on individual and community health needs the development of effective early diagnosis and treatment strategies. Suicide is one of the most concerning consequences of untreated depression, and it has been on the rise worldwide. Timely diagnosis and intervention are therefore crucial, not just to enhance patient outcomes, but also to reduce the rising suicide incidence. Traditional depression diagnosis relies mainly on self reported symptoms, structured interviews, and clinician based tests. While these procedures can be useful, they are susceptible to prejudice and variation in interpretation. More recently, developments in medical data analysis have centered on combining objective measures such as neuroimaging, biometrics, and electroencephalography (EEG). EEG, in example, provides a noninvasive, costeffective means of recording brain activity, which can then be examined to find patterns indicative of mental health problems such as depression. EEG data, which record electrical signals in the brain, shed light on the neural dynamics that may differ between healthy people and those suffering from depressive illnesses. With the rise of deep learning and its shown ability to handle complicated datasets, machine learning techniques are increasingly being used to medical signal data. Among these, Convolutional Neural Networks (CNNs) have demonstrated significant potential in processing structured data such as pictures and time series, making them ideal for EEG signal classification. Despite these advances, contemporary machine learning algorithms sometimes struggle with overfitting when applied to small or subjectspecific EEG datasets, resulting in lower generalizability across larger populations. To address these issues, this research offers DeprNet, a deep convolutional neural network framework developed specifically for classifying EEG data and distinguishing between depressed and nondepressed individuals. DeprNet is built with an emphasis on processing EEG signals obtained from both hemispheres of the brain, allowing for insights into the asymmetry of neural activity related with depression. The severity of depression in subjects is measured using the Patient Health Questionnaire 9 (PHQ9), a well regarded self assessment tool that ranks the intensity of depressed symptoms.
© 2024, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 513