ChestVision AI

🫁 AI system for chest disease diagnosis using both X-ray images and medical reports. The model combines: 📸 EfficientNetB3 for X-ray feature extraction 📝 BiLSTM for understanding medical reports After testing Early Fusion and Late Fusion, Hybrid Fusion with Attention Mechanism achieved the best performance by dynamically deciding whether to rely more on the image or the report for each case. ⚡ Challenges: * Severe data imbalance * Training on a personal laptop without GPUs Solutions included Class Weights and optimized training. 📊 Results: * 95–96% Weighted Accuracy * Pleural Effusion F1: 0.92 * Cardiomegaly F1: 0.93 * Pneumothorax Recall: 86–95% Grad-CAM was integrated for explainable AI visualization. 📚 TensorFlow · Keras · EfficientNetB3 · BiLSTM · Attention · Grad-CAM