Lung Segmentation using U-Net

This project implements a U-Net model to perform lung segmentation on medical images (e.g., CT scans or X-rays). With convolutional layers and skip connections, the U-Net model accurately segments lung regions, supporting various medical imaging tasks and disease analysis.

Features

• Custom Data Loader: Loads and preprocesses images and corresponding masks.
• Data Augmentation: Simple augmentations, such as horizontal flips, to increase data diversity.
• U-Net Architecture: Encoder-decoder structure with skip connections for detailed segmentation.
• Custom IoU Metric: Uses Intersection over Union (IoU) to evaluate segmentation accuracy.
• Training Callbacks: Includes model checkpointing, early stopping, and learning rate reduction.
• Visualization: Plots training history and displays sample predictions.

Project Structure

• DataLoader class: Loads images and masks, resizes to target size, and normalizes pixel values.
• SegmentationGenerator class: Generates batches of images and masks with optional augmentation.
• simple_unet function: Defines a U-Net model with an adjustable input size.
• iou_metric function: Calculates the IoU score as a custom evaluation metric.
• plot_training_history function: Plots training and validation loss and IoU metrics.
• predict_and_visualize function: Predicts masks on the validation set and visualizes results.

Getting Started

Prerequisites

• Python Libraries: Install the required libraries by running:

  pip install tensorflow numpy opencv-python scikit-learn matplotlib
  

Project Setup

Organize Data

Store images and corresponding mask files in separate directories:

dataset/ ├── images/ └── masks/

Adjust Paths

Update image_dir and mask_dir in the main script to point to your dataset directories.

Run the Project

To execute the project, run the main script. It will:

• Load and preprocess data
• Train the U-Net model with defined callbacks
• Save the best model
• Plot training metrics and visualize sample predictions

python lung_segmentation.py


## Results
-The script will output:

-Saved model files (best_model.keras and final_model.keras)
-Plots of training and validation metrics
-Sample predictions showing the original image, true mask, and predicted mask