Aigerim Djamanakova1, Andreia V. Faria2,
Kenichi Oishi2, Xin Li2, Kazi Akhter2,
Laurent Younes3,4, Peter van Zijl2,5, Michael Ira
Miller3, Susumu Mori2
1Biomedical Engineering, Johns Hopkins
School of Medicine, Baltimore, MD, United States; 2Radiology,
Johns Hopkins University, Baltimore, MD, United States; 3Center
for Imaging Science, The Johns Hopkins University, Baltimore, MD, United
States; 4Applied Mathematics & Statistics, Johns Hopkins
University, Baltimore, MD, United States; 5F.M. Kirby Center for
Functional Magnetic Resonance Imaging, Kennedy Krieger Institute, Baltimore,
MD, United States
We
used Large Deformation Diffeomorphic Metric Mapping in order to improve the
registration of brains with enlarged ventricles from patients with
Alzheimer's disease . By employing a
second channel of information comprised of the lateral ventricle segmentation
maps, obtained semi-automatically and automatically, we were able to increase
the accuracy of the mappings. The
degree of accuracy was calculated by comparing the results of the manual
segmentation of lateral ventricles and a neighboring structure, lingual
gyrus, with the single and dual-channel registration-based segmentation. This approach can be a powerful tool for
improving registration of images.