Diffusional Kurtosis along the Corticospinal Tract in Adult Normal Pressure Hydrocephalus

Demographics and Clinical Data

This retrospective, anonymized, single-center study was approved by the institutional review board with a waiver of consent and was Health Insurance Portability and Accountability Act–compliant. Demographics of the 3 patient groups (total n = 44) are given in Table 1. The 3 groups did not statistically differ in demographic characteristics.

Patients with NPH were selected from a pool of patients with enlarged ventricles and gait impairment referred to the New York University Adult Hydrocephalus Evaluation Program and evaluated by a neurologist (J.G.) with 25 years of experience. These patients all exhibited a characteristic dyspraxic gait, variable degrees of cognitive impairment, and symptoms of overactive bladder. From the initial chart review of patients with NPH from January 2003 through December 2014 (n = 624), we selected patients who fulfilled the following requirements: 1) completed a high-volume lumbar puncture or lumbar drain that resulted in convincing clinical improvement (clinician [J.G.] and the family's subjective impression of gait improvement), (n = 101); 2) the availability of preoperative 3T MR imaging acquired locally, which included a high-resolution MPRAGE sequence as well as diffusion images appropriate for diffusional kurtosis imaging analysis (n = 20); and 3) free of comorbidities such as cerebrovascular disease, coexisting intracranial mass lesions, or prior craniectomy identified by a neuroradiologist (A.G.). The final sample size was 14; however, 9 additional subjects scanned after December 2014 were identified from a review of 125 cases and were included in the study after having been found to meet criteria 2 and 3. Most excluded from the study failed to meet criterion 1. The full population of 23 subjects with clinical presentations of NPH evaluated in this study responded positively to shunt surgery as evaluated by a neurologist (J.G.). Of the 23 patients with hydrocephalus included in the study, 15 demonstrated a risk of hypertension, 7 patients had diabetes, 10 had hyperlipidemia, 3 had coronary artery disease, and 7 tested negative for any cardiovascular risk factors. Ten patients demonstrated comorbidities for ≥2 the risk factors listed above.

Control subjects were recruited retrospectively from the New York University Alzheimer Disease Center to undergo a full clinical research evaluation per Uniform Data Set procedures for Alzheimer Disease Centers,¹¹ and a brain MR imaging, including high-resolution MPRAGE and the appropriate diffusion MR imaging protocol. Healthy control subjects had no evidence of dementia or mild cognitive impairment and had a Clinical Dementia Rating global score of 0. Subjects with Alzheimer disease were given a diagnosis based on the Diagnostic and Statistical Manual of Mental Disorders¹² and the National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer Disease and Related Disorders Association criteria for probable Alzheimer disease¹³and a Clinical Dementia Rating of ≥0.5 (range, 0.5–2.0) and were not deemed to have any medical, neurologic, or psychiatric conditions that could otherwise account for the dementia. The 3 groups did not statistically differ in demographic characteristics (Table 1). Due to the sensitivity of the diffusion MR imaging metrics in general to age,¹⁴ age was included as a covariate in all analyses.

MR Imaging Acquisition and Image Processing

MR images were acquired on a 3T Tim Trio system (Siemens, Erlangen, Germany). Patients with NPH were scanned before ventricular shunt placement. The MR imaging protocol included 3D T1-weighted imaging for anatomic reference using an MPRAGE sequence and diffusion imaging with 3 b-values (0, 1000, 2000 s/mm²) along 60 diffusion-encoding directions using a single-shot twice-refocused echo-planar sequence. Data were acquired in an 88 × 80 × 30 matrix with 2.6 × 2.6 mm resolution and 5-mm slice thickness. Diffusion-weighted MR images were preprocessed by denoising using a Gaussian smoothing kernel, followed by motion and eddy current correction using the FMRIB Software Library (FSL; http://www.fmrib.ox.ac.uk/fsl) and then the kurtosis tensor was fitted using a constrained weighted linear least-squares fit using in-house software¹⁵ written in Matlab (MathWorks, Natick, Massachusetts). The kurtosis tensors were then used to derive the diffusional kurtosis imaging–acquired parametric maps of mean diffusivity, axial diffusivity (AD), radial diffusivity, fractional anisotropy (FA), mean kurtosis, axial kurtosis (AK), and radial kurtosis. The axonal water fraction (ie, the ratio of intra-axonal water to intra-plus-extra-axonal water) was computed in addition to standard diffusion and kurtosis measures based on a 2-compartment model of neuronal tissue.⁷

Along-Tract Analysis

Tracts were generated using MRtrix (github.com/MRtrix3/mrtrix3) by manually placing seed ROIs in the cerebral peduncles and in the precentral gyrus of each subject. Streamline propagation was constrained using an FA value of 0.2 and a maximum angle of 45°. Streamlines were reoriented to begin at the brain stem and stretch cephalad toward the precentral gyrus. They were then truncated at the middle cerebellar peduncle and pial surface to ensure that tracts from different subjects could be compared without individual anatomy biasing results. To prevent partial volume effects from potentially biasing the analysis, we excluded tracts that came close to or intersected with the lateral ventricles by creating an ROI of the lateral ventricles, dilating it by 1 voxel, and discarding any streamline that intersected this region.

Using Matlab, we normalized tracts using cubic spline interpolation so that the same anatomy occurred at the same distance along the tract for each subject, and vertices were relocated so that there were exactly 100 vertices per streamline along the entire tract, similar to the methods used in Colby et al.¹⁶ The cerebral peduncles were constrained to exist at 25% of the distance from the middle cerebellar peduncle, and the anterior end of the precentral gyrus was constrained to 95%. Parametric diffusion, kurtosis, and white matter tract integrity maps were resampled onto tract vertices as scalar values. Figure 1 provides a qualitative demonstration of axial diffusion and axial kurtosis mapped into the corticospinal tract of a representative sample of a patient with Alzheimer disease, a healthy control, and a patient with hydrocephalus.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Axial diffusion and axial kurtosis overlaid onto the corticospinal tract as a scalar value and represented by a color: Red indicates higher axial kurtosis or axial diffusivity, while blue indicates lower values. From left to right, we show sample tracts for patients with Alzheimer disease, healthy controls, and patients with hydrocephalus. Patients with NPH have much lower axial diffusion in periventricular regions than control groups and much greater axial kurtosis values in the same region next to the ventricles.

Periventricular white matter lesions were segmented for each subject based on T1 hypointensities and FLAIR hyperintensities from FreeSurfer parcellation of T1-weighted MPRAGE images (http://surfer.nmr.mgh.harvard.edu), followed by a nonlinear warp to a common space computed using the FMRIB Nonlinear Registration Tool (FNIRT; http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FNIRT). These segmentations were used to create leukoaraiosis probability maps to help visualize the location where this pathology is most likely to occur for patients with NPH, patients with Alzheimer disease, and healthy controls and to determine whether leukoaraiosis has a local impact on microstructural pathology associated with NPH.

Statistical Methods

One-way ANCOVA with age as a covariate was used to compare differences between patients with NPH and those with Alzheimer disease, those with NPH and healthy controls, and patients with Alzheimer disease and healthy controls at each point along the CST. A Bonferroni correction for multiple comparisons was used to correct for the presence of the 3 groups. The paired-samples t test was used to test for differences across hemispheres. Receiver operating characteristic curves were used to test the diagnostic accuracy using the area under the receiver operating characteristic curve (AUC) of each diffusion metric and to determine the ones to best differentiate among study groups. One-way ANCOVA controlling for age was used to test for group differences in periventricular leukoaraiosis.