1. email: vladimir.fonov@gmail.com Discussion & Conclusions
#452
Voxel-wise T2 relaxometry of Normal Pediatric Brain Development
Vladimir S. Fonov, Ilana R. Leppert, G. Bruce Pike, D. Louis Collins
McConnell Brain Imaging Centre, Montreal Neurological Institute, Montreal, QC, Canada
Results
Introduction
It is known that contrast between the grey and white matter in the MRI is reversed during the first 4-6 postnatal months. The actual timing of the change depends on the imaging sequence, field strength and the brain region. Here, data from an NIH-sponsored multi-center study of Normal Brain Development (NIH pediatric database [Almli 2007]) was used to perform automatic voxel-level analysis of the change of T2 relaxation over time for children 1-60 months old. The goal was to produce 3D relaxometry maps and quantify differences between lobes of the brain.
Out of 346 datasets, 326 have successfully passed all stages of automated processing, and resulting regressions show significant differences of change of T2 between different regions of the brain during the development.
Volumetric maps of A, B and C parameters are shown in Figure 3. Results of this study are consistent with a manual ROI based of T2-relaxometry performed on a subset of the same datasets in [Leppert 2008] ( Figure 4). Discrepancies (Major Forceps ROI) could be attributed to the difficulty of automatic segmentation of this anatomical location, especially on the youngest subjects.
Methods
MRI scans from 114 normal healthy children were acquired across 11 age cohorts [Almli 2007], with a total of 346 datasets ( Figure 1). Each dataset including one T1w scan (TR=500ms TE=12ms) and a dual echo T2w/PDw scan (TR=3500ms, TE=14,112ms), some datasets included an additional acquisition of dual echo scans with longer TE=83,165ms.
Figure 1. Subjects age distribution
Figure 3. Voxel-wise regression T2=A+B*exp(-C*t)
Figure 2. Age specific anatomical models (T1w)
Each dataset was processed in a following way. All dual echo scans were registered linearly to the T1w scan using a mutual information cost function. Individual T1w scans were nonlinearly registered to a custom average, age-specific pediatric template (Figure 2). The concatenated registration parameters were applied to the dual-echo scans to map all information from all subjects into the common coordinate system of the template. All images were manually inspected for the quality of registration. T2 was evaluated in each dataset in a voxel-wise manner using linearized equation: ln(Si)=ln(S0)-TEi/T2 , where S0 is the equilibrium signal and Si is the signal is corresponding echo time TEi).
Regressions of T2 over age was performed in a voxel-by-voxel fashion using all images that passed QC using mono-exponential (T2=A+B*exp(-C*t) ) model. Also, mean T2 values were extracted within several brain regions and regressions were performed on the extracted values.
Figure 4. Comparison of Manual and Automatic ROI T2 regressions
Discussion & Conclusions
We have shown an automated procedure to process relaxometry information in a large cohort of young subjects. Our results are consistent with the results of manual ROI based technique and allow to study the development of the brain in greater detail, showing quantitative differences in maturation of different parts of the brain.
Almli C.R, Rivkin M.J, McKinstry R.C. Brain Development Cooperative Group. The NIH MRI study of normal brain development (Objective-2): Newborns, infants, toddlers, and preschoolers. NeuroImage 2007:35: Clarke G.D. Phantom Test Guidance for the ACR MRI Accreditation Program and Phantom Site Scanning Instruction Guide. Virginia: American College of Radiology; p Leppert I.R., Almli C.R, McKinstry R.C. , Mulkern R.V., Pierpaoli C, Rivkin M. J., Pike G.B. T2 Relaxometry of Normal Pediatric Brain Development. Journal of Magnetic Resonance Imaging