YuraYara2005/CardioGuard-AI

CardioGuard AI is an enterprise-grade Clinical Decision Support System (CDSS) developed for the DEPI (Digital Egypt Pioneers Initiative) graduation project. # 🫀 CardioGuard AI: Real-Time Multi-Modal Cardiac Intelligence System CardioGuard AI is an enterprise-grade Clinical Decision Support System (CDSS) developed for the **DEPI (Digital Egypt Pioneers Initiative)** graduation project. The platform utilizes a high-throughput Apache Kafka-driven streaming pipeline capable of ingesting live ECG signals at 500Hz. These signals are processed using a cutting-edge hybrid deep learning architecture (**1D-CNN + TCN + Custom Attention**) to detect dangerous heart rhythm abnormalities in real time. To enhance clinical trust and bridge the gap between AI predictions and patient understanding, the system integrates Explainable AI (XAI) and a **Bilingual Generative AI scribe** to automatically translate diagnostic outputs into structured physician reports and patient-friendly Egyptian Arabic summaries. --- ## 🌟 Core Architecture & Features ### Big Data Streaming (Apache Kafka) - **High-Throughput Ingestion**: Sub-200ms latency ingestion of 12-lead ECG time-series data. - **Decoupled Architecture**: Resilient producer/consumer daemon architecture using `confluent-kafka` for Python. ### Hybrid Inference Engine (TensorFlow/Keras) - **Spatial & Temporal Analysis**: A **1D-CNN** backbone for spatial feature extraction combined with a **Temporal Convolutional Network (TCN)** and **Custom Attention** layers for deep temporal rhythm analysis. - **Medical Datasets**: Trained rigorously on the **PTB-XL** and **MIT-BIH Arrhythmia** datasets. ### Explainable AI (XAI) & Fusion - **Visual Heatmaps**: Integrates **Grad-CAM** and **SHAP** to generate visual heatmaps, highlighting the exact segments of the ECG signal that influenced the AI's prediction. - **Multi-Modal Fusion**: Combines physiological ECG signals with patient metadata (Age, BMI, Activity) for individualized baseline modeling. ### Data Synthesis (TimeSeriesGAN) - **Digital Twin Generation**: Addresses data imbalance in the MIT-BIH and PTB-XL datasets by generating high-f …