Up-sampling MRI scans can significantly increase robustness of NbM segmentation
Poster No:
315
Submission Type:
Abstract Submission
Authors:
Neda Shafiee1, D Louis Collins1
Institutions:
1McConnell Brain Imaging Centre, Montreal Neurological Institute, Montreal, Quebec
First Author:
Co-Author:
Introduction:
The main source of cholinergic projections to the cerebral cortex is the magnocellular neurons of the nucleus basalis of Meynert (NbM). Postmortem studies have shown high densities of Neurofibrillary tangles in NbM in early and presymptomatic stages of Alzheimer's disease. Degeneration of the cholinergic projection system has also been theorized to be an upstream event of entorhinal and neocortical degeneration, making NbM a possible biomarker early in the course of the disease. However, the precise delineation of NbM is difficult due to limited spatial resolution and contrast in MR images. The NbM also lacks strict boundaries with adjacent cell groups and different atlases used to identify NbM have reported discrepancy. To delineate this region more accurately, we propose to increase the resolution of MRI scans, before performing a deformation-based morphometry (DBM) analysis.
Methods:
MRI scans of 896 subjects from ADNI dataset (219 cognitively normal amyloid negative (CN-), 117 cognitively normal amyloid positive (CN+), 131 with early mild cognitive impairment (EMCI) and 242 with late MCI (LMCI) and 187 patients with AD) were up-sampled to 0.5 mm isotropic using a method introduced by Manjón et al. This method enforces a structure-preserving constraint as opposed to imposing an arbitrary smoothness constraint: the down-sampled version of the reconstructed image should be the same as the noise-free low-resolution image for all locations. The up-sampled scans were then non-linearly registered to an ADNI-based template. The resulting deformation fields were used to compute Jacobian determinant maps. All Jacobian maps were calculated in the template space, thus normalizing for head size. Using the atlas published by Zaborszky et al. as a mask, local volume change was computed for NbM. The same analysis was repeated but without up-sampling the data and using original 1 mm isotropic scans.
Results:
Comparing the results from high-resolution and original low-resolution scans shows that while both pipelines detect the general trend of NbM atrophy as the disease progresses, the high-resolution analysis finds a stronger distinction between disease stages. To quantify this distinction we used Cohen's d measurement and compared volume changes in early MCI against late MCI in both models. Upsampled scans showed a higher cohen's d and t-value, further confirming our assumption.
·early vs. late MCI differences for low-resolution and up-sampled analysis
·NbM volume change pairwise differences or low-resolution and up-sampled analysis
Conclusions:
When studying brain regions with small volumes, such as the Nucleus basalis of Meynert, enhancing the resolution would be beneficial to increase the measurement and segmentation accuracy.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Lifespan Development:
Aging
Modeling and Analysis Methods:
Methods Development
Segmentation and Parcellation 2
Keywords:
Aging
MRI
Segmentation
Other - Deformation-based morphometery, Nucleus basalis of Meynert
1|2Indicates the priority used for review
Provide references using author date format
Zaborszky, L. (2008), 'Stereotaxic probabilistic maps of the magnocellular cell groups in human basal forebrain', NeuroImage, vol. 42, pp. 1127-1141.