Imaging Methods Assessment and Reporting (IMAR)

The goal of the NLM's IMAR program is to provide data, methods, and computational resources for the quantitative comparison of image segmentation, registration, and computer-aided diagnosis methods.

This project is lead by Dr. Terry Yoo at the NLM and hosted by Kitware, Inc. using MIDAS, a collection of server/client tools for image and data assimilation.

Three types of experimental designs are supported for method evaluation. They are as follows:

1. Blinded Algorithm Evaluation

   Algorithm evaluators apply for access to evaluation data and upload their results. Results are automatically compared with sequested truth.
   • The Retrospective Image Registration Evaluation Project (RIRE) v2.0

2. Automated Evaluation

   Algorithm evaluators submit an executable to the website, and it is used to process sequestered data.
   • MIDAS: BatchMake (under construction)

3. Public Data Standards

   Algorithm evaluators are encouraged to download the data from these sites when conduction their own evaluations.
   • NLM's MIDAS Community
   • The BrainWeb project
     

Algorithm evaluators are encouraged to submit open-source implimentations and summaries of their results to the open-access Insight Journal.

New evaluation datasets and experiments are welcome. Please contact Dr. Terry Yoo or Dr. Stephen Aylward to arrange a submission.