Difference between revisions of "Projects:RegistrationDocumentation:ParameterTesting"

From NAMIC Wiki
Jump to: navigation, search
Line 8: Line 8:
 
*'''Objective:''' Determine optimized sets of parameters for successful automated registration of MR-MR images within the 3DSlicer software. DOF, cost function, initialization and optimization strategy will differ because of the differences in image contrast and/or content.  This work will present approaches and solutions for successful registration for a large set of combinations of MRI pairings. This is part of a concerted effort to build a Registration Case Library available to the medical imaging research community.
 
*'''Objective:''' Determine optimized sets of parameters for successful automated registration of MR-MR images within the 3DSlicer software. DOF, cost function, initialization and optimization strategy will differ because of the differences in image contrast and/or content.  This work will present approaches and solutions for successful registration for a large set of combinations of MRI pairings. This is part of a concerted effort to build a Registration Case Library available to the medical imaging research community.
 
*'''Method:'''
 
*'''Method:'''
−
**1- we choose 3-4 subjects/exams with 3-4 different contrast pairings: T1, T2, PD, FLAIR: ~12-16 images
+
**1- choose 3-4 subjects/exams with 3-4 different contrast pairings: T1, T2, PD, FLAIR: ~12-16 images. We choose sets for which we have a good registration solution
−
**2- we have an expert reader determine ~ 3-5 anatomical landmarks on each unregistered image
+
**2- disturb each pair by a known transform of varying rotational & translational misalignment
−
**3- we register all combinations and  run a sensitivity analysis for the most critical parameters: 6 vs. 12 DOF, cost function, % sampling
+
**3- run registration for a set of parameter settings and save the result Xform, e.g. metric: NormCorr vs. MI , 2% vs 5% sampling, 50 vs. 100 iteration max
−
**4- outcome metric is RMS error of fiducial alignment
+
**4- evaluate registration error as point distance and RMS. Plot error vs. initial misalignment (where does registration begin to fail), plot error vs. parameter settings (which setting works best for the toughest case)
−
**5- we report the best performing parameter set for each MR-MR combination
+
**5- run sensitivity analysis and report the best performing parameter set for each MR-MR combination
 
**6-extension 1: add different voxel sizes, i.e. emulate 1,3,5mm slice thickness
 
**6-extension 1: add different voxel sizes, i.e. emulate 1,3,5mm slice thickness
−
**7- extension 2: add initial misalignment as parameter to the test series
+
*This self-validation scheme avoids recruiting an expert reader to determine ~ 3-5 anatomical landmarks on each unregistered image pair (time constraint). Also we can cover a wider range of misalignment and sensitivity by controlling the input Xform. It also facilitates batch processing.
−
*Options: The expert landmark selection as rate-limiting step we could bypass by doing this as a self-validation where we start from a registration we consider optimal and then apply pre-determined misalignment. We then do not need fiducial pairs to evaluate but can derive RMS metrics from the result Xform directly.  We will have to justify/scrutinize how we chose our gold-standard. A true gold-standard would exist only for prospectively aligned image sets, such as a dual echo PD/T2.
 
−
*sensitivity analysis we report as line plots comparing RMS ranges for different metrics, e.g. compare MI vs. NCorr
 
  
 
[[Image:SRegTest_FLAIR-T1_LRRot15.png|left|500px|360 Level Tree]]
 
[[Image:SRegTest_FLAIR-T1_LRRot15.png|left|500px|360 Level Tree]]

Revision as of 21:10, 27 October 2009

Home < Projects:RegistrationDocumentation:ParameterTesting

Back to ARRA main page
Back to Registration main page
Back to Registration Use-case Inventory

ISMRM abstract 2010

  • Title: MR-protocol Tailored Medical Image Registration
  • Objective: Determine optimized sets of parameters for successful automated registration of MR-MR images within the 3DSlicer software. DOF, cost function, initialization and optimization strategy will differ because of the differences in image contrast and/or content. This work will present approaches and solutions for successful registration for a large set of combinations of MRI pairings. This is part of a concerted effort to build a Registration Case Library available to the medical imaging research community.
  • Method:
    • 1- choose 3-4 subjects/exams with 3-4 different contrast pairings: T1, T2, PD, FLAIR: ~12-16 images. We choose sets for which we have a good registration solution
    • 2- disturb each pair by a known transform of varying rotational & translational misalignment
    • 3- run registration for a set of parameter settings and save the result Xform, e.g. metric: NormCorr vs. MI , 2% vs 5% sampling, 50 vs. 100 iteration max
    • 4- evaluate registration error as point distance and RMS. Plot error vs. initial misalignment (where does registration begin to fail), plot error vs. parameter settings (which setting works best for the toughest case)
    • 5- run sensitivity analysis and report the best performing parameter set for each MR-MR combination
    • 6-extension 1: add different voxel sizes, i.e. emulate 1,3,5mm slice thickness
  • This self-validation scheme avoids recruiting an expert reader to determine ~ 3-5 anatomical landmarks on each unregistered image pair (time constraint). Also we can cover a wider range of misalignment and sensitivity by controlling the input Xform. It also facilitates batch processing.
360 Level Tree
Retrieved from "https://www.na-mic.org/w/index.php?title=Projects:RegistrationDocumentation:ParameterTesting&oldid=44396"