Difference between revisions of "2013 Project Week:BRAINSFit new features"
From NAMIC Wiki
| Line 2: | Line 2: | ||
<gallery> | <gallery> | ||
Image:PW-SLC2013.png|[[2013_Winter_Project_Week#Projects|Projects List]] | Image:PW-SLC2013.png|[[2013_Winter_Project_Week#Projects|Projects List]] | ||
| + | Image:SLC2013-Prostate_BRAINS_fixed.png|Reference test image | ||
| + | Image:SLC2013-Prostate_BRAINS_moving.png|Registered moving image | ||
</gallery> | </gallery> | ||
| Line 7: | Line 9: | ||
* BWH: Andrey Fedorov | * BWH: Andrey Fedorov | ||
| − | * Iowa: Hans Johnson | + | * Iowa: Hans Johnson, David Welch |
==Project Description== | ==Project Description== | ||
Latest revision as of 23:14, 10 January 2013
Home < 2013 Project Week:BRAINSFit new features
Reference test image
Registered moving image
Key Investigators
- BWH: Andrey Fedorov
- Iowa: Hans Johnson, David Welch
Project Description
Objective
- integrate, test and document new capabilities, which proved to be very helpful for prostate MRI registration, into BRAINSFit registration module
Approach, Plan
- specific features include:
- initializer based on the center of mask for the registration ROIs
- initialization of the B-spline transform support region restricted to the bounding box of the ROIs union
- ability to perform registration with only fixed ROI available
These features have been implemented earlier in a fork of ITK3-based BRAINSFit. We would like to merge these modifications with the main repository of BRAINS and test the capabilities with ITK4 on prostate MRI datasets.
Progress
- all features that we intended to integrate were implemented and merged with the BRAINS main repository (https://github.com/BRAINSia/BRAINSStandAlone/pull/160)
- in the process, several minor bugs were resolved
- test added and is passing on BRAINS dashboard
- Note: BRAINS is using ITK3 registration framework implemented in ITK4; it is currently not possible to evaluate the new registration framework of ITK4 with BRAINS
- Testing on 10 clinical datasets was performed using BRAINS+ITK4
- computation time is comparable with our in-house BRAINS implementation + ITK3
- registration was successful in 8 out of 10 cases, based on visual inspection; further investigation is needed for the failed cases and larger scale validation
