Rayz : AI-Powered Chest X-ray Analysis
🩺 Overview
This model analyzes chest X-rays to detect 14 potential lung conditions using EfficientNet_B0, a lightweight yet high-performing CNN. It was trained on NIH Chest X-ray Dataset & NLMCXR Dataset, providing reliable multi-class classification for various lung diseases.
🚀 Motivation
This project began when I received a false-positive tuberculosis (TB) report and had to wait for delayed X-ray results due to a holiday. Not knowing how to interpret X-rays, I built this AI tool to help others in similar situations.
📜 Model Details
- Model type: Image Classification (Chest X-ray Analysis)
- Architecture: EfficientNet_B0
- Trained on: NIH Chest X-ray & NLMCXR Datasets
- Input format: Chest X-ray images (
.png,.jpg) - Output: Probabilities for 14 lung conditions
- License: MIT
- Compute Requirement: Can run on CPU, optimized for GPU (CUDA)
💡 Why EfficientNet_B0?
I tested multiple models, including DenseNet121, ViT, and CNNs, but EfficientNet_B0_best_93.44 outperformed the others in terms of:
- High Accuracy (AUROC: 0.72 - 0.93)
- Lower Computational Cost
- Faster Inference Speed
- Better Generalization across datasets
📊 Model Performance
| Model | AUROC Score (Avg) |
|---|---|
| EfficientNet_B0 | 0.72 - 0.93 |
| DenseNet121 | 0.55 - 0.95 |
| ViT_Base | 0.32 - 0.65 |
🔧 How to Use the Model
1️⃣ Install Dependencies
pip install torch torchvision transformers pillow numpy matplotlib seaborn
Inference Providers
NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API:
The model has no library tag.