Published August 1, 2018 | Version 1.0
Other Open

VGG Lung Nodule Classifier

  • 1. University of California, Los Angeles

Description

VGG Lung Nodule Classification Algorithm
Shiwen Shen, William Hsu

This algorithm classifies whether a 2D image patch contains a nodule or non-nodule. The algorithm is distributed as a Docker module. This release requires access to an NVIDIA GPU.

To run the Docker module,
 1. Install NVIDIA Docker
    https://github.com/NVIDIA/nvidia-docker
 2. Unzip archive
 3. Build Docker module (example command below for Ubuntu 16.04)
    sudo nvidia-docker build -t vgg-nodule-classifier -f Dockerfile.gpu .
 4. Run Docker module
    sudo nvidia-docker run vgg-nodule-classifier

When the Docker module executes, the entire workflow from model training to evaluation will be executed with the final model outputted as the file vgg_model_and_weights.h5.

To transfer this file out of the Docker module, use the following command:

 docker cp vgg-nodule-classifier:/vgg_model_and_weights.h5 <dest_path>

Files

vgg-nodule-classifier.zip

Files (381.8 MB)

Name Size
md5:3d2e38012e6fe5aab1eec91590e37cde
381.8 MB Preview Download