Precision of template-based inter-subject spatial normalization of QSM of the older adult brains

2214 

Abstract Submission 

Rasheed Abid¹, Yingjuan Wu², Mohammad Niaz², Abdur Raquib Ridwan², Arnold Evia², David Bennett², Konstantinos Arfanakis^1,2

¹Illinois Institute of Technology, Chicago, IL, ²Rush University Medical Center, Chicago, IL

Rasheed Abid  
Illinois Institute of Technology
Chicago, IL

Yingjuan Wu  
Rush University Medical Center
Chicago, IL

Mohammad Niaz  
Rush University Medical Center
Chicago, IL

Abdur Raquib Ridwan  
Rush University Medical Center
Chicago, IL

Arnold Evia  
Rush University Medical Center
Chicago, IL

David Bennett  
Rush University Medical Center
Chicago, IL

Konstantinos Arfanakis  
Illinois Institute of Technology|Rush University Medical Center
Chicago, IL|Chicago, IL

Quantitative Susceptibility Mapping (QSM) holds significant potential for studying metal and iron homeostasis. It can serve as an important diagnostic tool for various pathologies [1]. However, there are no studies on the precision of QSM spatial normalization for older adults, which is contingent upon the quality and representativeness of the chosen template, as well as the type and quality of information utilized during image registration. This study aims to compare three available QSM templates in terms of their representativeness of the older adult brain, and in terms of the precision of inter-subject matching of older adult QSM data when they are used as references for spatial normalization.

Data:
In this study, 3D T1-weighted MPRAGE and multi-echo 3D GRE data from 100 older adults (aged 67.8-97.2 years) were used. Recently introduced T1w and QSM [2-4] templates of the MIITRA atlas were compared to the only two other sets of publicly available T1w and QSM templates from HybraPD [5] and MuSus-100 [6], mainly based on younger adults.

Process:
Magnetic Susceptibility Map Generation: For each participant, magnetic susceptibility maps were generated using the Morphology Enabled Dipole Inversion (MEDI) software on multi-echo GRE data [7].
Image Registration: Five distinct registration pipelines were employed using the ANTS image registration toolbox [8]:
- Single channel registration of T1-weighted images to T1-weighted templates.
- Multi-channel registration of T1-weighted images and susceptibility maps to T1-weighted (75% weight) and QSM templates (25% weight).
- Equal weight multi-channel registration of T1-weighted images and susceptibility maps to both T1-weighted and QSM templates.
- Multi-channel registration of T1-weighted images and susceptibility maps to T1-weighted (25% weight) and QSM templates (75% weight).
- Single channel registration of susceptibility maps to QSM templates.
Deformation Calculation: Log-Jacobian maps of the nonlinear deformations were generated for each participant for registration to each space, and the average Log-Jacobian was calculated for each voxel in each space.
Template Alignment: The transformations from the image registration step were used to align the participants' susceptibility maps to the different atlases.
Spatial Matching Precision: Inter-subject pairwise normalized cross-correlation was calculated for each registration approach and each atlas, considering all 100 participants (100x99/2=4950 pairs).

Spatial normalization of older adult brains required less deformation when registering to the recently generated MIITRA templates compared to other templates (Fig.1). This was true for all five registration approaches tested here. Thus, the MIITRA templates were more representative of the older adult brain. Furthermore, pairwise normalized cross-correlation was higher when using single channel registration of susceptibility maps to QSM templates than other registration methods (Fig.2). Overall the highest spatial matching of magnetic susceptibility maps of older adults was achieved when using single channel registration of magnetic susceptibility maps to the MIITRA-QSM template.

This study provides insights into the role of different atlases and registration approaches in spatial alignment of magnetic susceptibility maps of the older adult brain. It demonstrates that the recently developed MIITRA atlas is more representative of the older adult brain, as evidenced by the lower amount of deformation required compared to other atlases. The present study also shows that the highest inter-subject spatial matching of magnetic susceptibility maps from older adult brains is achieved when using single channel registration of individual magnetic susceptibility maps to the MIITRA QSM template. These findings underscore the importance of selecting appropriate templates and registration pipelines for QSM studies of the older adult brain.

Aging ²

Brain Atlases ¹

Aging

Atlasing

Data Registration

Spatial Normalization