MRI-based quantification of cardiac-driven brain biomechanics for early detection of neurological disorders

We present a pipeline to quantify biomechanical environment of the brain using solely MRI-derived data in order to elucidate the role of biomechanical factors in neurodegenerative disorders. Neurological disorders, like Alzheimer’s and Parkinson’s diseases, are associated with physical changes, including the accumulation of amyloid-β and tau proteins, damage to the cerebral vasculature, hypertension, atrophy of the cortical gray matter, and lesions of the periventricular white matter. Alterations in the external mechanical environment of cells can trigger pathological processes, and it is known that AD causes reduced stiffness in the brain tissue during degeneration. However, there appears to be a significant lag time between microscale changes and macroscale obstruction of neurological function in the brain. Here, we present a pipeline to quantify the whole brain biomechanical environment to bridge the gap in understanding how underlying brain changes affect macroscale brain biomechanics. This pipeline enables image-based quantification of subject-specific displacement field of the whole brain to subject-specific strain, strain rate, and stress across 133 labeled functional brain regions. We have focused our development efforts on utilizing solely MRI-derived data to facilitate clinical applicability of our approach and have emphasized automation in all aspects of our methods to reduce operator dependance. Our pipeline has the potential to improve early detection of neurological disorders and facilitate the identification of disease before widespread, irreversible damage has occurred.

isotropic, compressible Mooney-Rivlin hyperelastic constitutive law () [40,41] for both white matter and gray matter in the present study with material constants provided below in Equation A7Error!Reference source not found..The cerebrospinal fluid domain has been neglected presently while focusing on the regions of white matter lesion formation in the brain tissue.
Eqn. A7 Where  is the isotropic, compressible Mooney-Rivlin estimation of strain energy density formulated comparable to Giordano et al. 2017[42], G and K are the shear and bulk modulus respectively,  is the determinant of the deformation gradient,  1 ̃ is the first invariant of the isochoric Cauchy-Green strain tensor.
Following the calculation of strain energy density (Wiso), the first Piola-Kirchhoff stress tensor (P), and Cauchy stress tensor (σ) can be estimated by calculating the partial derivative of the strain energy density at each spatial location with respect to the deformation gradient as shown in Equations A8-A9, where  is the Jacobian of the deformation gradient.

B: Subject-specific strain and stress value breakdown
Although the subject population used in the present study contains no medical information, it may be valuable for researchers to understand the expected scale of strain and stresses that are estimated by this pipeline and its assumptions, namely the maximum displacement prescribed and the simplistic Mooney-Rivlin model.Thus, the output of the automated pipeline is reported in Table 2 below with mean values, standard deviations, and maximums shown for each functional brain region (in Alzheimer's Disease related regions of interests).Additionally, simple group-level averaging is reported with mean, standard deviations, and maximum values shown.Means were averaged across subjects, standard deviations were computed using the weighted average of the squared standard deviations, and maximums were computed as the maximum value across subjects.

C: Image processing
The structural MRI and cine MRI scans are evaluated on different grid orientations due to a quirk of the amplified MRI processing algorithm.Specifically, the displacement field (stored as a 5D MATLAB data structure)is not aligned with the T1-weighted structural MRI.Thus, a pipeline was generated to re-align these images for use in generating automated segmentations.The following steps detail the default settings applied to this pipeline, all of which have been scripted to be performed automatically upon launching the driving script and can be found in the raw2ondisp.m file.
Note: The grid transformations are only required when the subject's T1-weighted MRI and displacement field are on different grids or orientations, which may not be the case in all applications. C.1.

Table B1 :
Summary table of von-Mises strain and stress across subjects and functional brain regions associated with AD.Values are shown as:

Invert the T2_to_T1.mat transform to get a mapping from the T1 grid & orientation to the DispField grid and orientation (same as T2 grid and orientation)
o FSL FLIRT XFM -Invert FLIRT transform ▪ Transformation Matrix for A to B: T2_to_T1.mat▪ Save Inverse Transorm (B to A): T1_to_T2.matC.5.

Generate 133-label SLANT Segmentation on T1 grid
o NOTE: Directory needs to be adjusted before running SLANT as SLANT will automatically evaluate all .nii.gz files in the provided domain ▪ Make a folder "T1w_scan" in the patient directory export input_dir=/home/tdiorio/Documents/0_DATA_UWash_AMRI/S11/T1w/ ▪ sudo mkdir $input_dir ▪ export output_dir=$input_dir/output ▪ sudo nvidia-docker run -it --rm -v $input_dir:/INPUTS/ -v $output_dir:/OUTPUTS masidocker/public:deep_brain_seg_v1_1_0 /extra/run_deep_brain_seg.sh ▪ CAREFUL: The /home/input_dir is just an example location, you will need to specify this path to reflect the location of the T1-weighted scan you are looking to process.▪ Note: SLANT will generate a 133-label segmentation for each compatible .nii.gz file in the $input_dir.Following the aforementioned naming conventions will allow the AMRI2Stress pipeline to automatically choose the correct SLANT segmentation during violin plot generation."./T1w/output/FinalResult/S11_T1w_MPR_BIC_SLANT.nii.gz"C.9.