2010 Summer Project Week EMSegmentation kmeans
Key Investigators
- BWH: Priya Srinivasan, Daniel Haehn, Sylvain Bouix
- UPenn: Kilian Pohl
Objective
To implement as one of the options: KMeans automatic initialization of tissue class intensities in the EMSegmentation Module in Slicer3.
Approach, Plan
Currently we have stand alone python scripts that take in 3 inputs: a template MRML scene, path to the T1 image, path to the atlas images. Intensities are computed using KMeans and a MRML scene is outputted which can directly be loaded in Slicer3 for segmentation. Registration of images (T1 and atlases) has to be done as a preprocessing step before providing as inputs to the python scripts. Meet up during NAMIC week and see how best this process can be integrated as one of the options in the current EMSegmentation pipeline in Slicer3.
Progress
Met with Kilian and went over the current processing pipeline with the stand alone Python codes and now have a general understanding of how we should proceed in order to integrate into Slicer 3.6. Will work out the details in the coming weeks. Based on Kilian's suggestions, modified the Python codes to make it more easy to integrate into Slicer 3.6.
Delivery Mechanism
This work will be delivered to the NA-MIC Kit as a Loadable module.
References:
- A Hierarchical Algorithm for MR Brain Image Parcellation. K. Pohl, S. Bouix, M. Nakamura, T. Rohlfing, R. McCarley, R. Kikinis, W. Grimson, M. E. Shenton, W. Wells. IEEE Transactions on Medical Imaging. 2007
