Chest_XRay_Classifier_App_README.md

Chest X-Ray Pathology Classifier — Streamlit Application

A production-ready web application for AI-assisted chest X-ray pathology detection, built with PyTorch and Streamlit.

---

Overview

This application enables users to upload a chest X-ray image and receive instant AI-powered predictions for 14 common thoracic pathologies. The underlying model is a DenseNet-121 trained via transfer learning on the NIH ChestX-ray14 dataset.

Pathology	Description
Atelectasis	Collapsed lung tissue
Cardiomegaly	Enlarged heart
Consolidation	Fluid-filled lung tissue
Edema	Fluid accumulation
Effusion	Fluid around lungs
Emphysema	Damaged air sacs
Fibrosis	Scarred lung tissue
Hernia	Diaphragmatic hernia
Infiltration	Substance accumulation
Mass	Abnormal growth
Nodule	Small abnormal growth
Pleural Thickening	Thickened pleura
Pneumonia	Lung infection
Pneumothorax	Collapsed lung (air leak)

Clinical Disclaimer: This tool is for educational and research purposes only. It is NOT a substitute for professional medical diagnosis. Always consult a qualified radiologist or physician.

---

Live Demo

Access the deployed application: chest-xray-classifier-app-tc.streamlit.app

[App Screenshot with an sample Chest X-Ray]

---

Project Structure

Chest-XRay-Classifier-App/
├── app.py                 # Main Streamlit application
├── model.py               # Model loading and inference logic
├── utils.py               # Image preprocessing utilities
├── requirements.txt       # Python dependencies
├── Dockerfile             # Container configuration
├── .streamlit/
│   └── config.toml        # Streamlit UI configuration
├── assets/
│   ├── app_screenshot.png # Demo screenshot
│   └── sample_xrays/      # Example images for testing
├── models/
│   └── densenet121_chestxray.pth  # Pre-trained weights (not in repo)
└── README.md

---

Technology Stack

Layer	Technology
Frontend	Streamlit
Deep Learning	PyTorch, TorchVision
Model	DenseNet-121 (Transfer Learning)
Image Processing	PIL, OpenCV
Visualization	Matplotlib, Plotly
Deployment	Streamlit Cloud / Docker

---

Quick Start

1. Clone the Repository

git clone https://github.com/tishact7/Chest-XRay-Classifier-App.git
cd Chest-XRay-Classifier-App

2. Install Dependencies

pip install -r requirements.txt

3. Download Model Weights

The pre-trained model weights are hosted via GitHub Releases due to file size constraints. Download the latest release and place the weights file in:

models/densenet121_chestxray.pth

Alternatively, train your own model using the classifier training repository.

4. Run the Application

streamlit run app.py

The application will open at http://localhost:8501.

---

Docker Deployment

# Build the image
docker build -t chest-xray-app .

# Run the container
docker run -p 8501:8501 chest-xray-app

---

Model Performance

The underlying model was trained on the NIH ChestX-ray14 dataset (112,120 images, 30,805 unique patients):

Metric	Value
AUC-ROC (Macro)	~0.82
AUC-ROC (Micro)	~0.90
Input Size	224 x 224 px
Architecture	DenseNet-121
Classes	14 (multi-label)

Full training details and reproducible notebooks are available in the Chest-XRay-Pathology-Classifier repository.

---

Usage

1. Upload a chest X-ray image (PNG or JPG supported).
2. View the preprocessed image.
3. Review predicted probability scores for all 14 pathologies.
4. Interpret results using the built-in visualization.

Supported Image Formats

• .png, .jpg, .jpeg

---

Configuration

Edit .streamlit/config.toml to customize the UI:

[theme]
primaryColor = "#FF4B4B"
backgroundColor = "#0E1117"
secondaryBackgroundColor = "#262730"
textColor = "#FAFAFA"
font = "sans serif"

---

Sample Output

{
  "Atelectasis": 0.87,
  "Cardiomegaly": 0.12,
  "Consolidation": 0.45,
  "Edema": 0.03,
  "Effusion": 0.91,
  "Emphysema": 0.08,
  "Fibrosis": 0.15,
  "Hernia": 0.01,
  "Infiltration": 0.67,
  "Mass": 0.22,
  "Nodule": 0.34,
  "Pleural_Thickening": 0.19,
  "Pneumonia": 0.55,
  "Pneumothorax": 0.04
}

---

Future Enhancements

• Grad-CAM visualization to highlight model attention regions
• Batch upload for processing multiple images simultaneously
• PDF report generation for clinical documentation
• DICOM metadata display for patient/study information
• Model comparison toggle (DenseNet, ResNet, EfficientNet)
• User-adjustable confidence thresholds

---

About the Author

Tisha Chatterjee is a Biomedical Engineering student and independent researcher converging on brain-computer interfaces, neuroplasticity, and medical imaging. She is a researcher at MIT Critical Data, a Global Neurosurgery Fellow at George Washington University, a cohort member at Stanford-affiliated S.Y.A.L.I.S Labs (top 8% of 2,000+ applicants), and a bioinformatics researcher at CureQuest. She is a TEDx Scholar (Top 10 globally) and an IIT Bombay Techfest College Ambassador (Rank 36 / 10,000+).

• 🔗 LinkedIn
• 📧 tishachatterjee77@gmail.com
• 🧠 Research Interests: BCI · Neuroplasticity · Medical Imaging · Computational Biology · AI in Healthcare

---

Citation

If you use this work in your research, please cite:

@misc{chatterjee2026chestxray,
  author = {Chatterjee, Tisha},
  title = {Chest X-Ray Classifier App: Multi-Label Thoracic Disease Detection with DenseNet-121},
  year = {2026},
  howpublished = {\url{[https://github.com/tishact7/Chest-XRay-Classifier-App]}},
  note = {Independent research project}
}

Acknowledgments

• NIH ChestX-ray14 Dataset — National Institutes of Health
• PyTorch and TorchVision — Meta AI Research
• Streamlit — For streamlined ML deployment

---

Contact

For questions, collaborations, or feedback:

• Open an Issue
• Connect on LinkedIn

---

AI in healthcare should augment, not replace, clinical expertise.