hacsmith/lung-nodule-seg

Lung Nodule Segmentation with Attention UNet

This space segments lung nodules from CT scan slices using an Attention UNet model.

Features

• —🖼️ Multiple format support: PNG, JPG, JPEG, BMP, TIFF
• —📊 MHD medical format: Directly process .mhd files
• —🎯 Automatic slice selection: For 3D volumes
• —🏥 CT windowing: Optimized lung window settings
• —📈 Interactive visualization: Overlay and metrics

Usage

1. 1.Wait for the model to load (first time may take a moment)
2. 2.Upload a CT image or MHD file
3. 3.Adjust the threshold if needed
4. 4.Click "Segment Nodules"
5. 5.View results in the tabs

Model Architecture

• —Type: Attention U-Net
• —Input: 1-channel grayscale (512×512)
• —Output: Binary segmentation mask
• —Parameters: ~8 million
• —Training: LUNA16 dataset

Technical Details

• —MHD files are processed with SimpleITK
• —CT windowing: Center=40, Width=400 (lung window)
• —Default slice: Middle slice of 3D volume
• —Image resizing: 512×512 pixels

File Structure

======= title: Lung Nodule Seg emoji: 📚 colorFrom: gray colorTo: indigo sdk: gradio sdkversion: 6.6.0 appfile: app.py pinned: false ---

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

>>>>>> d68378300ed3f09e7be710b83044cb68ac902683