brain imaging studies: high throughput by …t1 12m t1 06m t1 24m t2 24m t1 step 2 register the 6...
TRANSCRIPT
RESEARCH POSTER PRESENTATION DESIGN © 2011
www.PosterPresentations.com
Brain Imaging Studies: High throughput by Automatic Processing Pipeline
We have collected 711 neonate brain MR images from an ongoing autism study (Autism Centers of Excellence - Infant Brain Imaging Study ACE-IBIS) . Each infant gets 3 scans at 6month, 1year and 2 years of age in order to characterize brain growth trajetories.
Since the same child gets repeated scans over time, we can combine the image processing of all three time points into a joint analysis step, which increases robustness and improves consistency of measures over the age range.
We have developed a fully automatic image processing pipeline to register serial scans of each subject, to provide tissue and brain lobe segmentation, and to segment major subcortical structures.
INTRODUCTION
OBJECTIVES
Longitudinal Registration Pipeline
METHODS METHODS RESULTS
CONCLUSION
The pipeline presented was used to parallel process all 711 cases on a 251 core HPC for several days and calculated the volume of WM, GM, CSF , left and right hemispheres, parcellation, and sub-cortical structures in mm3. The segmentation results were also used in computing the cortical thickness on WM and GM. In the future, we will model the intensity changes over time and use this model to further refine the segmentation and registration.
REFERENCE[1], Marcel Prastawa, John Gilmore, Weili Lin, and Guido Gerig. Automatic Segmentation of Neonatal Brain MRI. Medical Image Computing and Computer Assisted Intervention (MICCAI) 2004. Lecture Notes in Computer Science (LNCS) 3216, Pages 10-17.[2], D. Rueckert, L. I. Sonoda, C. Hayes, D. L. G. Hill, M. O. Leach, and D. J. Hawkes. Non- rigid registration using free-form deformations: Application to breast MR images. IEEE Transactions on Medical Imaging, 18(8):712-721, 1999.[3], J. A. Schnabel, D. Rueckert, M. Quist, J. M. Blackall, A. D. Castellano Smith, T. Hartkens, G. P. Penney, W. A. Hall, H. Liu, C. L. Truwit, F. A. Gerritsen, D. L. G. Hill, and D. J. Hawkes. A generic framework for non-rigid registration based on non-uniform multi-level free-form deformations. In Fourth Int. Conf. on Medical Image Computing and Computer- Assisted Intervention (MICCAI '01), pages 573-581, Utrecht, NL, October 2001.[4], C. Studholme, D.L.G.Hill, D.J. Hawkes, An Overlap Invariant Entropy Measure of 3D Medical Image Alignment, Pattern Recognition, Vol. 32(1), Jan 1999, pp 71-86.
1. Register the images scanned at 6, 12 and 24 months all to 24 months for all 711 cases.
2. Use this longitudinal registration information for segmentation:
1. White matter, Gray matter, CSF2. Hemisphere segmentation3. Brain parcellation segmentation4. Subcortical segmentation
3. Volume computing on based on the segmentation results listed above.
4. Cortical thickness computations on white matter and gray matter.
5. Segmentation of a set of subcortical structures.
SCI Institute, University of Utah
Xiaoyue Huang, Sylvain Gouttard, Guido Gerig
06MT1
06MT2
12MT1
12MT2
24MT1
24MT2
06MT1
06MT2
12MT1
12MT2
24MT1
24MT2
Step1 Bias correction and EMS segmentation
EMS segmentation & bias correction
24MT2
12MT2
06MT2
24MT1
12MT1
06MT1
24MT2
24MT1
Step 2 Register the 6 and 12 month T2 images to the 24 month T2 image
Linear registration + nonlinear registration
Step 3 Register the 24 month T1 image to the 24 month T2 image
Step 4 Register the 6 and 12 month T1 images to 24 month T1 image
Linear registration + nonlinear registration
Linear registration + nonlinear registration
Hemisphere and Parcellation Segmentation
24MT2
24MT1
24Mlabel
06Mseg
Nonlinear transformation to produce the12 and 06 month labels
12Mseg
24Mskullstrip
24MHemisphere
atlas
Nonlinear registration
12Mlabel
06Mlabel
06Mmask
12Mmask
The images above show the longitude registration result for one case on T1 modality. The images obtained at different times are transformed to the same position and size, enabling us to compare the intensity changes at a particular position.
Subcortical Segmentation and Cortical Thickness
(left column from top) Subcortical segmentation results for one case at 6, 12 and 24 months.(right column from top) Gray matter thickness for one case at 6, 12 and 24 months. The green to red color transitions represent the cortical thickness from thin to thick.
(left column from top) Hemisphere segmentation results for one case at 6, 12 and 24 months.(right column from top) Parcellation labels for 6, 12 and 24 month scans.
(left column from top) By using the processing pipeline above, we can get the WM (red) and GM (green) segmentation results for one case at 6, 12 and 24 months.(right column from top) Registered T2 images obtained at 6, 12 and 24 months.
White and Gray Matter Segmentation
Color bar measured in mm
T1:
T2:
www.sci.utah.edu
SCI
INST
ITUTE ● EXHIBIT ● EXPLORE ● EXC
ITE EXPERIENCE ● E
XCHAN
GE ●
SCI