README.md

metadata

title: Chest X Ray Disease Classifier
emoji: ⚕️
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 3.48.0
app_file: app.py
pinned: false
license: mit
tags:
  - medical-imaging
  - computer-vision
  - multi-label-classification
  - chest-xray
  - deep-learning
  - tensorflow
  - efficientnet
short_description: 15 thoracic diseases detection (AUC 0.784)

Chest X-Ray Disease Classification 🏥

Automated detection of 15 thoracic diseases from chest X-ray images using deep learning.

🎯 Performance

• Mean AUC: 0.784 (beats 2017 baseline by +5.9%)
• Recall: 80.3% (medical priority - catch diseases early)
• Architecture: EfficientNetB0 with full fine-tuning
• Dataset: NIH ChestX-ray14 (112,120 images)

🔬 Model Details

Training:

• Focal Loss for class imbalance
• Balanced sampling (oversampling rare diseases)
• Test-Time Augmentation (TTA)
• Mixed Precision (FP16)
• Patient-level train/test split

Best Performing Diseases:

• Edema: 0.884 AUC
• Cardiomegaly: 0.865 AUC
• Effusion: 0.852 AUC

🔥 NEW! Grad-CAM Visualization:

• Enable the checkbox to see where the model looks
• Red regions = High attention (important for prediction)
• Blue regions = Low attention (ignored by model)
• Helps validate the model isn't using spurious features

⚠️ Limitations

IMPORTANT: This is a research prototype. NOT for clinical diagnosis.

• High false positive rate (60%) by design to maximize recall
• Dataset has label noise (NLP-extracted from reports)
• Single-site training (NIH) - may not generalize
• Requires radiologist review for all predictions
• NOT FDA-approved or clinically validated

📊 Use Case

Intended: Screening tool to flag suspicious X-rays for radiologist review
NOT intended: Standalone diagnosis, emergency triage, legal liability scenarios

🔗 Resources

• Dataset: NIH ChestX-ray14 on Kaggle
• Code: GitHub Repository
• Paper: Wang et al. 2017 - ChestX-ray14 Dataset

📄 Citation

@article{wang2017chestxray14,
  title={Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases},
  author={Wang, Xiaosong and Peng, Yifan and Lu, Le and Lu, Zhiyong and Bagheri, Mohammadhadi and Summers, Ronald M},
  journal={Proceedings of the IEEE conference on computer vision and pattern recognition},
  pages={2097--2106},
  year={2017}
}

---

Author: Emir Muhammet Aran | Institution: Computer Engineering Student
Last Updated: December 2025