Expert System for Detection and Grading of Depression in Youngsters

Project Description

This project presents a multi-modal expert system designed to detect and assess depression levels in young individuals, using advanced AI techniques and a three-tiered user structure: Patient, Doctor, and Admin.

The system leverages real-time data through a combination of:

The depression detection model is built using Convolutional Neural Networks (CNN) and Bidirectional LSTM, trained on the FER-2013 dataset and refined using insights from the DAIC-WOZ corpus. A guided triggering video is used to evoke emotional responses, followed by verbal feedback analysis, offering a more holistic approach to mental health evaluation.

Multi-User Architecture

Patient Side

  • Take PHQ-9 depression test + video/audio input analysis
  • View test results and depression severity grading
  • Book appointments with available doctors
  • Receive digital prescriptions
  • Pay consultation fees via integrated billing

Doctor Side

  • View upcoming appointments
  • Access patient test results and emotional analysis
  • Track billing and payment status
  • Manage individual patient records
  • Prescribe medicine directly through the system

Admin Panel

  • Full system control and monitoring
  • Manage users, doctors, appointments, and analytics
  • Oversee testing data, billing, and prescription history

Key Features

Real-time Analysis

Comprehensive depression detection using facial, audio, and questionnaire inputs

AI-Powered

Advanced analysis with CNN + BiLSTM models for accurate assessment

Emotion AI

Sentiment and emotion recognition pipeline for deeper insights

Smart Prescriptions

Dynamic prescription system tailored to patient needs

Workflow Automation

Seamless end-to-end appointment and billing workflow

Secure Access

Role-based access control ensuring security and privacy

Impact

This project combines AI, psychology, and system design to provide a practical tool for early detection of depression in educational and clinical environments, improving both accuracy and accessibility over traditional self-assessment methods.