Luke Bloy1, Alex R. Smith2, Madhura Ingalhalikar2, Robert T. Schultz3, Timothy P.L. Roberts4, Ragini Verma2
1Section of Biomedical Imaging, University of Pennsylvania, Philadelphia, PA, United States; 2Section of Biomedical Imaging, Univeristy of Pennsylvania, Philadelphia, PA, United States; 3Center for Autism Research, Children's Hospital of Philadelphia, Philadelphia, PA, United States; 4Lurie Family Foundation's MEG Imaging Center, Children's Hospital of Philadelphia, Philadelphia, PA, United States
This study compares the results of using DTI and HARDI based diffusion models as the driving force behind spatial normalization algorithms. Each modality underwent separate state of the art registration pipelines designed to optimally take advantage of each contrast. The deformations resulting from each pipeline were applied to the images of the other modality, allowing for three means of comparison. Both registration pipelines perform similarly when FA variance was used as a means of comparison, however using either FOD or normalized FOD variance HARDI registration performed better. This demonstrates the importance of using HARDI when accurate registration is required.