---
language: en
tags:
- deep-learning
- computer-vision
- medical-imaging
- chest-xray
- pneumonia
- keras
- tensorflow
license: cc-by-4.0
datasets:
- chest-xray-pneumonia
widget:
- text: Upload a chest X-ray image (.jpg or .png)
---

# 🩻 Pneumonia Detection Model (CNN)

## 🧠 Overview
This model is a **Convolutional Neural Network (CNN)** trained to classify **chest X-ray images** as either **Normal** or **Pneumonia**.  
It demonstrates the application of **deep learning** to **medical imaging classification** and serves as an **educational and research tool** — **not** for clinical use.

---

## ⚙️ Model Details

- **Architecture:** Custom CNN (3 convolutional + 2 dense layers)  
- **Framework:** TensorFlow / Keras  
- **Input size:** 150×150 RGB images  
- **Output:** Binary classification — `0 = Normal`, `1 = Pneumonia`  
- **Dataset:** *Chest X-Ray Images (Pneumonia)* — [Kermany et al., *Cell* (2018)](http://www.cell.com/cell/fulltext/S0092-8674(18)30154-5)

---

## 📚 Dataset Information

**Source:**  
Kermany, D.S., Goldbaum, M., Cai, W. *et al.*  
*Identifying Medical Diagnoses and Treatable Diseases by Image-Based Deep Learning*,  
*Cell* (2018), 172(5), 1122–1131.e9  
[DOI: 10.1016/j.cell.2018.02.010](https://doi.org/10.1016/j.cell.2018.02.010)

**License:** [Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/)

---

## 🩺 Intended Use

This model is designed for:
- Machine learning and data science education  
- Research in medical image analysis  
- Prototyping of AI healthcare applications  

⚠️ **Disclaimer:**  
This model is **not a medical device** and **should not be used for clinical or diagnostic purposes**.  
It is provided **solely for educational and research use**.

---

## 🧩 How It Works

1. The user uploads a **chest X-ray image** (`.jpg` or `.png`).
2. The image is resized to **150×150** pixels and normalized to `[0, 1]`.
3. The CNN processes the image and outputs a probability:
   - `0 → Normal`
   - `1 → Pneumonia`
4. The final classification is the class with the highest confidence score.

---

## 📊 Evaluation Report

### 🧾 Classification Metrics
| Metric | Normal | Pneumonia |
|:--------|:--------:|:-----------:|
| **Precision** | 0.962 | 0.829 |
| **Recall** | 0.659 | 0.985 |
| **F1-score** | 0.783 | 0.900 |
| **Support** | 232 | 389 |

**Overall accuracy:** 0.863  
**Macro F1-score:** 0.841  
**Weighted F1-score:** 0.856  

---

### 📉 ROC Curve  
![ROC Curve](https://huggingface.co/dlopez350/pneumonia_detector/resolve/main/roc_curve.png)

**AUC = 0.959**

---

### 📊 Confusion Matrix  
![Confusion Matrix](https://huggingface.co/dlopez350/pneumonia_detector/resolve/main/confusion_matrix.png)

| True Label | Predicted Normal | Predicted Pneumonia |
|-------------|------------------|---------------------|
| **Normal** | 153 | 79 |
| **Pneumonia** | 6 | 383 |

---

## 💡 Example Usage

```python
import tensorflow as tf
import numpy as np
from tensorflow.keras.preprocessing import image

# Load model
model = tf.keras.models.load_model("cnn_pneumonia.keras")

# Load and preprocess image
img_path = "test_image.jpg"
img = image.load_img(img_path, target_size=(150, 150))
img_array = image.img_to_array(img) / 255.0
img_array = np.expand_dims(img_array, axis=0)

# Predict
prediction = model.predict(img_array)[0][0]
label = "Pneumonia" if prediction > 0.5 else "Normal"
confidence = prediction if prediction > 0.5 else 1 - prediction

print(f"Prediction: {label} ({confidence:.2f} confidence)")
```