Detection & Classification of Cardiac Arrhythmia

An AI-powered diagnostic tool for real-time ECG analysis and clinical reporting

---

📝 Project Overview

This repository implements a state-of-the-art ECG Arrhythmia Detection and Classification system. Designed for clinicians and medical researchers, it bridges the gap between raw signal data and actionable medical insights using advanced Machine Learning.

🌟 Key Features

Feature	Description
🔍 Multi-Source Input	Support for ECG CSVs, high-res images, and manual physiological data.
🧠 Weighted KNN Intelligence	Classifies 16 arrhythmia types with high precision using a pre-trained Weighted KNN model.
📊 Signal Visualization	Interactive Plotly dashboards for wave analysis (P, QRS, T waves).
📄 Clinical Reports	Automated PDF generation with patient data and risk assessment.
👨‍⚕️ Doctor Portal	Dedicated dashboard for patient management and appointment scheduling.
🔔 Instant Alerts	Email notifications for high-risk cardiac events.

---

🏥 Clinical Workflow

1. Patient Intake: Register patient data and record vital signs.
2. ECG Analysis: Upload raw ECG data (image/CSV) or enter manual readings.
3. Model Inference: AI extracts features and predicts the arrhythmia category.
4. Review & Report: Doctor reviews the findings and generates a clinical PDF.
5. Follow-up: System triggers alerts to relevant medical staff if necessary.

---

🎨 Methodology & System Architecture

📊 Methodology Flowchart

graph LR
    subgraph Data_Acquisition ["📥 Data Acquisition"]
        A[ECG Signal CSV]
        B[ECG Image]
        C[Manual Vitals]
    end

    subgraph Preprocessing ["⚙️ Preprocessing & Feature Extraction"]
        D{Input Type?}
        E[Signal Filtering & Noise Removal]
        F[Image Thresholding & Segmenting]
        G[Feature Vector Generation]
    end

    subgraph AI_Engine ["🧠 AI Inference Engine"]
        H[Weighted KNN Classifier]
        I[Arrhythmia Type Prediction]
        J[Risk Level Assessment]
    end

    subgraph Clinical_Output ["🏥 Clinical Output"]
        K[(SQLite Databases)]
        L[PDF Medical Report]
        M[Email Alerts]
    end

    A & B & C -.-> D
    D -.->|Signal| E
    D -.->|Image| F
    D -.->|Manual| G
    E & F -.-> G
    G -.-> H
    H -.-> I -.-> J
    J -.-> K & L & M

    linkStyle default stroke:#333,stroke-width:2px,stroke-dasharray: 5;
    style Data_Acquisition fill:#e1f5fe,stroke:#01579b,stroke-width:2px
    style Preprocessing fill:#fff3e0,stroke:#e65100,stroke-width:2px
    style AI_Engine fill:#f3e5f5,stroke:#4a148c,stroke-width:2px
    style Clinical_Output fill:#fce4ec,stroke:#880e4f,stroke-width:2px

Loading

🏗️ System Architecture

graph LR
    subgraph Client ["🌐 Frontend"]
        UI[Browser Interface]
        Plotly[Plotly Viewer]
    end

    subgraph Server ["🚀 Backend (Flask)"]
        API[API Routes]
        Auth[Auth/Session]
    end

    subgraph Logic ["🧠 Processing & ML"]
        PP[Preprocessing]
        KNN[Weighted KNN]
        PDF[PDF Engine]
    end

    subgraph Data ["💾 Data & Storage"]
        DB[(SQL Databases)]
        FS[Asset Storage]
    end

    subgraph Notify ["📩 Notifications"]
        
        Mail[SMTP Email]
    end

    UI <-.-> API
    API -.-> Auth
    API -.-> PP
    PP -.-> KNN
    KNN -.-> PDF
    API -.-> DB
    API -.-> FS
    PDF -.-> FS
    API -.-> Notify

    linkStyle default stroke:#555,stroke-width:2px,stroke-dasharray: 3;
    style Client fill:#e8f5e9,stroke:#2e7d32
    style Server fill:#fffde7,stroke:#fbc02d
    style Logic fill:#efebe9,stroke:#4e342e
    style Data fill:#f3e5f5,stroke:#7b1fa2
    style Notify fill:#e0f7fa,stroke:#006064

Loading

---

🛠️ Technology Stack

• Backend: Flask (Python 3.8+)
• Database: SQLite3 (Distributed hospital.db and user_data.db)
• Machine Learning: Scikit-Learn (Weighted KNN), NumPy, SciPy, Pandas
• Image Analysis: OpenCV, Matplotlib
• Frontend: HTML5, CSS3, JavaScript (Plotly.js for signals)
• Reporting: ReportLab (Professional PDF Engine)
• Communication: SMTP (Email)

---

📂 Project Structure

• app.py: Main Flask application containing all routes and logic.
• NOTEBOOK_FILES/: Contains the pre-trained model.pkl.
• datasets/ecg_images/: Directory for ECG image data.
• static/: Assets including CSS, images, and generated reports.
• templates/: Jinja2 templates for the clinical web interface.
• requirements.txt: Python dependencies.- hospital.db: SQLite database for hospital/patient records.- user_data.db: SQLite database for user authentication data.- .env.example: Template for environment variables (API keys, secrets).- avg.csv / avgg.py: ECG feature averaging utilities.- calibrate_ecg_from_kaggle.py: Script to calibrate ECG data from Kaggle datasets.- hospital.db: SQLite database for hospital/patient records.- user_data.db: SQLite database for user authentication data.- .env.example: Template for environment variables (API keys, secrets).- hospital.db: SQLite database for hospital/patient records.

---

🚀 Future Roadmap

• ☁️ Cloud Integration: Sync patient reports with Azure/AWS Health.
• 📱 Mobile App: Flutter-based companion for real-time patient monitoring.
• ⏱️ Real-time IoT: Support for live ECG streaming from wearable sensors.
• 🤖 LLM Summaries: Automated clinical notes generation using Gemini/GPT-4.

---

Developed with ❤️ by Jaideep M C and Dhanush T
Empowering Healthcare with Intelligent Machine Learning