NeuroPredict: Deep Learning and LLM-Based Framework for SSRI Response in Major Depressive Disorder

Md Fahimul Kabir Chowdhury, Jayed Mohammad Barek, Md Ariful Hasan, Haili Wang

Keywords: SSRI, Major Depressive Disorder, CNN, Mental Depression, Machine Learning

Abstract: Major Depressive Disorder (MDD) is a leading cause of disability worldwide, yet selective serotonin reuptake inhibitors (SSRIs) yield highly variable outcomes across patients. To address the need for individualized treatment prediction, we introduce NeuroPredict, a non‑invasive EEG‑based framework that leverages rich time–frequency representations of pre‑treatment signals. EEG data were rigorously preprocessed with artifact removal and transformed using Wigner-Ville distribution (WVD), continuous wavelet transform (CWT), and constant‑Q transform (CQT), then classified with a tailored convolutional neural network (CNN). We systematically benchmarked NeuroPredict against state‑of‑the‑art pretrained models (Xception, DenseNet201, MobileNetV2) and EEG‑specific baselines (EEGNet, SleepEEGNet, DeepConvNet). Our proposed CNN consistently outperformed these approaches, achieving a peak accuracy of 98.28% and ROC of 98.20%, alongside strong precision, recall, and F1 score. Beyond predictive gains, we also highlight how large language model (LLM)-based support can enhance interpretability and streamline the analytic pipeline. These findings establish NeuroPredict as a scalable and clinically viable tool for precision prediction of SSRI treatment response, advancing the integration of data‑driven methods into personalized psychiatry.

Primary Subject Area: Transfer Learning and Domain Adaptation

Secondary Subject Area: Application: Neuroimaging

Registration Requirement: Yes

Reproducibility: https://github.com/fahimulkabir/NeuroPredict

Visa & Travel: Yes

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Submission Number: 260