NA-MIC Project Weeks

Back to Projects List

Deep Learning Integration in Slicer

Key Investigators

• Jorge Onieva (BWH)
• Raúl San José (BWH)
• Roya Khajavi
• Alireza Mehrtash
• Andrey Fedorov

Project Description

Integrate a lung segmentation algorithm based on Deep Learning in Slicer.

Objective

1. Develop a proof of concept to test the integration of Keras+Tensorflow tools in Slicer
2. Create a Slicer package that can be distributed with these features

Approach and Plan

1. Integrate the algorithm+pretrained models in CIP (see CIPDeepLearningLungSegmentation).
2. Compile Slicer against a customized Python that includes all the CIP required components
3. Pack Slicer with that Python version

Progress and Next Steps

1. This integration was done through the CustomSlicerGenerator in MacOS and Linux.
2. Luckily, it would be obsolete in Slicer 5!! A template with a Python distribution based on Anaconda or others may be used
3. Also, we found out other extensions like DeepInfer and TOOMCAT that may be useful in the meantime