Axial Diffusivity of the Corona Radiata Correlated With Ventricular Size in Adult Hydrocephalus

Transcription

1 Neuroradiology/Head and Neck Imaging Original Research Cauley and Cataltepe Imaging Adult Hydrocephalus Neuroradiology/Head and Neck Imaging Original Research Keith A Cauley Oguz Cataltepe 2 Cauley KA, Cataltepe O Keywords: axial diffusivity, corona radiata, diffusiontensor imaging, hydrocephalus, radial diffusivity DOI:24/AJR Received September 9, 202; accepted after revision November 5, 203. This project was supported by a grant from the University of Massachusetts Department of Radiology. Department of Radiology, Division of Neuroradiology, Columbia Presbyterian Medical Center, 630 W 68th St, New York, NY Address correspondence to K. A. Cauley (keithcauley@hotmail.com). 2 Department of Neurosurgery, UMass/Memorial Medical Center, University of Massachusetts, Worcester, MA. AJR 204; 203: X/4/ American Roentgen Ray Society Axial Diffusivity of the Corona Radiata Correlated With Ventricular Size in Adult Hydrocephalus OBJECTIVE. Hydrocephalus causes changes in the diffusion-tensor properties of periventricular white matter. Understanding the nature of these changes may aid in the diagnosis and treatment planning of this relatively common neurologic condition. Because ventricular size is a common measure of the severity of hydrocephalus, we hypothesized that a quantitative correlation could be made between the ventricular size and diffusion-tensor changes in the periventricular corona radiata. In this article, we investigated this relationship in adult patients with hydrocephalus and in healthy adult subjects. MATERIALS AND METHODS. Diffusion-tensor imaging metrics of the corona radiata were correlated with ventricular size in 4 adult patients with acute hydrocephalus, 6 patients with long-standing hydrocephalus, and 48 consecutive healthy adult subjects. Regression analysis was performed to investigate the relationship between ventricular size and the diffusion-tensor metrics of the corona radiata. Subject age was analyzed as a covariable. RESULTS. There is a linear correlation between fractional anisotropy of the corona radiata and ventricular size in acute hydrocephalus (r = 84, p < 0.00), with positive correlation with axial diffusivity (r = 36, p = 0.04) and negative correlation with radial diffusivity (r = 68, p = 0.009). In healthy subjects, axial diffusion in the periventricular corona radiata is more strongly correlated with ventricular size than with patient age (r = 66, p < 0.00, compared with r = 0.058, p = 69). CONCLUSION. Axial diffusivity of the corona radiata is linearly correlated with ventricular size in healthy adults and in patients with hydrocephalus. Radial diffusivity of the corona radiata decreases linearly with ventricular size in acute hydrocephalus but is not significantly correlated with ventricular size in healthy subjects or in patients with long-standing hydrocephalus. V entricular enlargement is seen at imaging of hydrocephalus but is also seen in cerebral atrophy or can be a normal variant. Hydrocephalus can be communicating or arise from a CSF flow obstruction. Advances in MRI may serve to increase the diagnostic accuracy of MRI of hydrocephalus, improve our ability to determine the type or cause of hydrocephalus, and aid in treatment planning for this relatively common neurologic condition. Diffusion-tensor imaging (DTI) metrics, such as fractional anisotropy and mean diffusivity, as well as axial and radial diffusivities offer means of characterizing normal and abnormal water diffusion in brain tissue. Previous studies have evaluated the white matter (WM) tracts adjacent to the lateral ventricles in various forms of hydrocephalus and found that the WM fibers of the corona radiata show increased fractional anisotropy, with increased axial diffusion in pediatric obstructive hydrocephalus as well as in normal-pressure hydrocephalus, when compared with age-matched control subjects []. This observation may have direct clinical relevance for normal-pressure hydrocephalus because the fibers of the corona radiata include descending motor fibers and gait abnormality is one of the hallmarks of this condition [, 2]. More acute forms of hydrocephalus present with symptoms of elevated intracranial pressure, such as headache, rather than with the Hakim triad of gait abnormality, urinary incontinence, and dementia that are characteristic of normal-pressure hydrocephalus. In all cases of hydrocephalus, the imaging diagnosis is based on ventricular size, and the severity of hydrocephalus and of the clinical findings is thought to correlate with the degree of ventricular enlargement. 70 AJR:203, July 204

2 Imaging Adult Hydrocephalus Because previous studies have focused on the WM fibers of the corona radiata and because these fibers may have direct clinical relevance in this condition, we chose this WM region as the topic for our study. We hypothesized that the diffusion-tensor characteristics of the WM fibers of the corona radiata are proportional to the degree of ventricular enlargement and that the relationship between various diffusion tensor metrics of the corona radiata and ventricular size may vary with the type of hydrocephalus. For example, longstanding hydrocephalus might be expected to show greater WM damage and therefore increased mean diffusivity when compared with more acute forms of hydrocephalus or to healthy control subjects. We arbitrarily addressed two classes of hydrocephalus in this TABLE : Hydrocephalus Patient Data and Clinical Information Patient No. Age (y) Sex Cause Surgery Acute hydrocephalus 47 F Colloid cyst Cyst resection 2 48 F Colloid cyst Cyst resection 3 23 M Brain tumor Endoscopic third ventriculostomy 4 42 M Aqueductal stenosis Endoscopic third ventriculostomy 5 72 F Aqueductal stenosis Endoscopic third ventriculostomy 6 68 M Brain tumor Endoscopic third ventriculostomy 7 50 F Brain tumor Endoscopic third ventriculostomy 8 22 F Aqueductal stenosis Endoscopic third ventriculostomy 9 40 F Colloid cyst Endoscopic third ventriculostomy 0 53 M Brain tumor Ventriculoperitoneal shunt 45 F Hematoma Endoventricular drain 2 39 F Aqueductal stenosis Endoscopic third ventriculostomy 3 28 M Brain tumor Ventriculoperitoneal shunt 4 8 M Brain tumor Ventriculoperitoneal shunt Long-standing hydrocephalus 50 F Arrest hydrocephalus Endoscopic third ventriculostomy 2 8 F Arrest hydrocephalus Endoscopic third ventriculostomy 3 29 F Arrest hydrocephalus Endoscopic third ventriculostomy 4 39 F Arachnoid cyst third ventricle Endoscopic third ventriculostomy study: acute obstructive hydrocephalus and long-standing hydrocephalus of any cause. We compared our findings with those from a large number of control subjects. Our study elaborates on the alterations in the diffusion tensor that occur with increased ventricular size in hydrocephalus. In this study we showed a linear increase in fractional anisotropy and axial diffusivity and a linear decrease in 5 32 F Arrest hydrocephalus Endoscopic third ventriculostomy 6 48 F Arrest hydrocephalus Endoscopic third ventriculostomy 7 5 F Arrest hydrocephalus Endoscopic third ventriculostomy 8 58 F Arrest hydrocephalus Endoscopic third ventriculostomy 9 64 F Colloid cyst Endoscopic third ventriculostomy/ventriculoperitoneal shunt 0 68 F Normal-pressure hydrocephalus Ventriculoperitoneal shunt 48 M Arrest hydrocephalus None 2 70 M Probable normal-pressure hydrocephalus None 3 65 M Probable normal-pressure hydrocephalus None 4 8 M Normal-pressure hydrocephalus Ventriculoperitoneal shunt 5 75 F Normal-pressure hydrocephalus Ventriculoperitoneal shunt 6 73 M Normal-pressure hydrocephalus Ventriculoperitoneal shunt Note Acute hydrocephalus indicates patients who had sudden onset of symptoms without previous hydrocephalus and who needed urgent or emergent intervention. Long-standing hydrocephalus indicates patients asymptomatic or symptomatic more than 6 months with previously documented hydrocephalus. Arrest hydrocephalus indicates patients with macrocephaly and severely enlarged ventricles and with no symptoms or with symptoms for more than 6 months. Probable normal-pressure hydrocephalus indicates patients with typical normal-pressure hydrocephalus symptoms [24] and good response to large-volume CSF removal but who did not have ventriculoperitoneal shunt. Normal-pressure hydrocephalus indicates patients with typical symptoms who had symptomatic improvement after ventriculoperitoneal shunt placement. AJR:203, July 204 7

3 Cauley and Cataltepe radial diffusivity within the WM fibers of the corona radiata as a function of ventricular size in acute hydrocephalus. We found a significant increase only in axial diffusivity as ventricular size increases in chronic hydrocephalus. We also found that axial diffusivity increases with ventricular size in healthy subjects, and this correlation is significantly greater than the correlation with subject age. Materials and Methods Study Population The study was approved by our institutional committee on human research. All routine brain studies include a 5-direction DTI acquisition with the exception of cases with specifically requested protocols or with metallic interference from shunt or previous surgery. A waiver of informed consent was granted by the institutional review board at our institution for this study. All studies were acquired from a single MR scanner (Signa, GE Healthcare) over a 3-year period from August 2009 to August 202. Because inclusion of pediatric cases could confound the study over questions regarding incomplete myelination, our focus was on the diffusion characteristics of the periventricular WM in adult patients with hydrocephalus. Consecutive cases of hydrocephalus were identified through the neurosurgery department. Inclusion criteria were the following: 8 years old or older and MRI study of the brain that included the DTI sequence of sufficient quality for analysis. This resulted in 34 total cases. These cases were then divided into classes of acute hydrocephalus and longstanding hydrocephalus. The study population is summarized in Table. Acute hydrocephalus cases present with acute clinical signs of increased intracranial pressure with imaging evidence of enlarged ventricles and a point of CSF flow obstruction. Cases of significant leukoaraiosis were excluded because WM disease may represent a confounding variable (one case). Cases with typical appearance of transependymal capping (six cases) seen in acute hydrocephalus were not excluded. For acute hydrocephalus, the patient age ranged from 8 to 72 years, with mean age ± SD of 42.5 ± 6. years (six men and eight women). Long-standing hydrocephalus patients have known hydrocephalus of long duration or present with chronic symptoms. Normal-pressure hydrocephalus cases were included in this group because there were insufficient numbers of cases to analyze separately. Cases with surgery or a shunt at the level of the corona radiata, which significantly impacts image quality, were excluded (three cases). Cases of significant leukoaraiosis were excluded because WM disease may represent a confounding variable (one case). For longstanding hydrocephalus, the patient ages ranged from 8 to 8 years, with a mean age of 54.3 ± 8.45 years (five men and women). Inclusion criteria for control subjects were the following: a normal brain MRI, no known neurologic diagnosis, and no focal neurologic symptoms. Forty-eight consecutive patients meeting these criteria were included (34 women and 4 men), with patient ages ranging from 9 to 83 years, with a mean age of 43.8 ± 4.3 years. Indications for brain MRI were headache (n = 34), weakness (n = 5), vertigo (n = 6), dizziness (n = 2), and syncope (n = ). Imaging Parameters All imaging was performed using a standard head coil. DTI was performed in the axial plane using single-shot echo-planar imaging with the following parameters: TR/TE, 8000/83.2; diffusion gradient encoding in 5 directions; b = 0 and 000 s/mm 2 ; FOV, mm; matrix size, 28 28; section thickness, 5 mm; and number of signals averaged,. A total of 26 slices without interslice gap was used to cover the cerebral hemispheres, upper brainstem, and cerebellum. A parallel imaging technique (Array Spatial Sensitivity Encoding Technique [ ASSET], GE Healthcare) was used to decrease the diamagnetic susceptibility artifact over the petrous bones acceleration factor (R = 2) during DTI acquisition. MRI brain slices are positioned parallel to the bicommissural line in standard clinical fashion. As part of our standard brain imaging protocol, each patient underwent unenhanced sagittal T- weighted imaging and transverse T2-weighted imaging. The sequence parameters for T2-weighted imaging were as follows: TR/TE, 2600/04; FOV, 24 cm; matrix size, 384 (frequency direction) 224 (phase direction); slice thickness, 5 mm; slice gap,.0 mm; and one signal acquired. Image Postprocessing Ventricular size was assessed using the frontal occipital horn ratio, which is the ratio of the width of the frontal horns plus the width of the occipital horns divided by twice the interparietal diameter. This technique has been shown to be a reliable method for estimating ventricular volume [3]. More strictly, our interest is with ventricular size measured on 2D (single-slice) images because the diffusion region of ROI is placed on single images and hydrocephalus is typically assessed radiographically on the basis of ventricular size rather than ventricular volume. Ventricular measurements were made on axial T2-weighted images using electronic calipers on the PACS workstation. Fig. Axial color-coded fractional anisotropy map image in 42-year-old man with obstructive hydrocephalus (patient 4) shows ROI placement in corona radiata (rectangles). DTI postprocessing analysis was performed with DTI fiber tracking software, release.9.0 (Medinria) [4]. For computation of diffusion tensors, raw data were transferred onto a workstation containing the Medinria software and processed. After generation of the diffusion-tensor matrix from the diffusion-weighted series, the three eigenvalues or diffusivities (λ, λ 2, and λ 3 ) were calculated by matrix diagonalization [5, 6]. The mean diffusivity is the average of the three eigenvalues and of the diffusion tensor, and it measures the magnitude of water molecule diffusion. Fractional anisotropy is defined as a coefficient of variation of the eigenvalues and is an index of the degree of directionality of water diffusivity [7, 8]. In the CNS, λ (axial diffusivity) represents the water diffusivity parallel to the axonal fibers. The average of the lesser eigenvalues [(λ 2 + λ 3 ) / 2] is the radial diffusivity, perpendicular to the axonal fibers [7, 8]. A single ROI method was used to define the ascending fibers of the corona radiata as has been described by others []. The corona radiata was identified on directionally color-coded fractional anisotropy maps, and rectangular freehand ROIs were placed over descending fibers of the bilateral corona radiata at the midventricular level (Fig. ). All MRI analysis was performed by a single experienced observer who was blinded to subject identity. The midportion of the corona radiata was chosen, and the ROI was drawn to include the entire width of the corona radiata. At this level, the genu of the corpus callosum and transverse fibers of the ascending longitudinal fasciculus are typically seen [9]. Left and right hemispheric corona radiata values were averaged. 72 AJR:203, July 204

4 Imaging Adult Hydrocephalus TABLE 2: Summary of Single-Variable Linear Regression Correlation Data Variable r p 95% CI r p 95% CI Acute hydrocephalus Corona radiata fractional anisotropy 89 < to Mean diffusivity to to Axial diffusivity to.00 Radial diffusivity to 8 Frontal occipital horn ratio to 9 Long-standing hydrocephalus Corona radiata fractional anisotropy to to 95 Mean diffusivity to to 80 Axial diffusivity to to 45 Radial diffusivity to Frontal occipital horn ratio Control subjects Corona radiata fractional anisotropy Mean diffusivity Axial diffusivity Radial diffusivity 46 < Frontal occipital horn ratio Fractional Anisotropy Radial Diffusivity 0.8 y =.49x r 2 = 226 A y = 0.908x r 2 = 367 D Data Analysis The main aim of the analysis was to correlate the diffusion characteristics of the corona radiata with the ventricular size estimated by the frontal occipital horn ratio across the three populations: control subjects, patients with acute hydrocephalus, and patients with long-standing hydrocephalus. We correlated the diffusion characteristics of the corona radiata with the ventricular size estimated by the frontal occipital horn ratio. Linear regression analysis was performed with derivation of the correlation coefficient (r) and the correlation of determination (r 2 ) as well as significance (p). Scatterplots and regression lines were drawn in Microsoft Excel 200, and statistics (r, r 2, and p) were derived or confirmed ADC (mm 2 /s) y = 69x r 2 = B Age in IBM SPSS, version 20) statistical software, with significance at p < Separate hierarchic regression analysis was performed for each of the groups in which the variable age was used as a covariate. Frontal occipital horn ratio was entered as the independent variable, and the various diffusion characteristics of the corona radiata were accordingly entered as dependent Axial Diffusivity y =.5726x r 2 = 408 C Fig. 2 Diffusion characteristics of corona radiata in adult patients with acute hydrocephalus. A, Scatterplot shows fractional anisotropy values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (coefficient of determination [r 2 ] = 2, p < 0.00). B, Scatterplot shows mean diffusivity (apparent diffusion coefficient [ADC]) values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (r 2 = 0.08, p = 55). C, Scatterplot shows axial diffusivity values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (r 2 = 4, p = 0.04). D, Scatterplot shows radial diffusivity values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (r 2 = 4, p = 0.009). AJR:203, July

5 Cauley and Cataltepe Fractional Anisotropy 0.8 y = 654x r 2 = A variables for the run of the regression. In step of the hierarchic regression, age was entered as a predictor. In step 2, the frontal occipital horn ratio was entered as a predictor. This two-step process allowed us to assess the effect of frontal occipital horn ratio on the diffusion characteristics over and above the effect due to age. To enable us to compare the regression models across the three groups, we used the Fisher z test to assess the fit between the model structures. The bivariate zero-order correlations between variables and semipartial correlations between frontal occipital horn ratio and diffusion characteristics were accordingly generated. ADC (mm 2 /s) B Analyses were done using the IBM SPSS statistics software, and some of the computations for the standard errors (SEs) were manually calculated. The z test used to compare the differences in regression weights across the three groups was based on the following equation: SE b difference = SE 2 + SE 2 2 and z = y = 2.642x r 2 = 65 Fig. 3 Diffusion characteristics of corona radiata in adult patients with long-standing hydrocephalus. A, Scatterplot shows fractional anisotropy values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (coefficient of determination [r 2 ] = 0.036, p = 83). B, Scatterplot shows mean diffusivity (apparent diffusion coefficient [ADC]) values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (r 2 = 6, p = 9). C, Scatterplot shows axial diffusivity values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (r 2 = 7, p = 0.03). D, Scatterplot shows radial diffusivity values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression(r 2 = 0.023, p = 79). Fractional Anistropy 0.8 y = 538x r 2 = A b b 2 SE 2 + SE 2 2 where b is the mean value of the given group. Results Analyses of the single-factor scatterplots discussed in this article are summarized in Table 2. ADC (mm 2 /s) y = x r 2 = B Fig. 4 Diffusion characteristics of corona radiata as function of ventricular size in control subjects. A, Scatterplot shows fractional anisotropy values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (coefficient of determination [r 2 ] = 0.034, p = 0.072). B, Scatterplot shows mean diffusivity (apparent diffusion coefficient [ADC]) values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (r 2 = 0.080, p = 0.09). C, Scatterplot shows axial diffusivity values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (r 2 = 0.99, p < 0.00). Polynomial curve fit yields slightly higher r 2 value and takes into account stronger correlation at frontal occipital horn ratio greater than 50. D, Scatterplot shows radial diffusivity values of corona radiata compared with ventricular size estimated with frontal occipital horn ratio with linear regression (r 2 = 0.005, p > 5). Axial Diffusivity Radial Diffusivity y =.8843x r 2 = 69 C 0.9 y = 549x r 2 = D Diffusion-Tensor Characteristics of Corona Radiata in Acute Hydrocephalus Correlation between ventricular size and fractional anisotropy and mean diffusivity Fractional anisotropy of the corona radiata were measured for each subject and plotted against the frontal occipital horn ratio. As the scatterplot (Fig. 2A) shows, there is a significant positive correlation between fractional anisotropy of the corona radiata and ventricular size in hydrocephalus (r = 89, r 2 = 23, p < 0.00). Mean diffusivity of the corona radiata was also computed for each study and plotted Axial Diffusivity Radial Diffusivity y = 27.46x x r 2 = 389 y =.98x + 97 r 2 = C y = 022x r 2 = D 74 AJR:203, July 204

6 Imaging Adult Hydrocephalus TABLE 3: Summary of Regression Across Groups Variables against the frontal occipital horn ratio. The scatterplot (Fig. 2B) shows no significant correlation between mean diffusivity and ventricular size in adult hydrocephalus (r = 6; r 2 = 0.08, which was not significant; p = 55). Correlation between ventricular size and axial and radial diffusivity The principal eigenvalue (axial diffusivity) of the corona radiata was calculated for each study and plotted against the frontal occipital horn ratio. The scatterplot (Fig. 2C) illustrates a strong positive correlation between axial diffusivity and the ventricular size in hydrocephalus (r = 63, r 2 = 4, p = 0.04). The radial diffusivity of the corona radiata was calculated for each study and plotted against the frontal occipital horn ratio. The scatterplot (Fig. 2D) shows a strong negative correlation between ventricular volume and radial diffusivity measurements within the corona radiata in acute adult hydrocephalus (r = 6, r 2 = 37, p = 0.009). Correlations between diffusivity measurements of corona radiata and patient age To determine the correlation between diffusivity metrics and patient age in acute hydrocephalus, we plotted patient age against the fractional anisotropy, mean diffusivity, and axial and radial diffusivity values of the corona radiata. There was little correlation between these variables in adults with acute hydrocephalus: age and corona radiata fractional anisotropy (r = 68, r 2 = 0.072, p = 55), age and corona radiata mean diffusivity (r = 42, r 2 =, p = 32), age and corona radiata radial diffusivity (r = 33, r 2 = 0.87, p = 2), and age and corona radiata axial diffusivity (r = 0.0, r 2 < 0.00, p = 0.97). Diffusion-Tensor Characteristics of Corona Radiata in Long-Standing Hydrocephalus Correlation between ventricular size and fractional anisotropy and mean diffusivity Fractional anisotropy of the corona radiata was measured for each study and plotted against the frontal-occipital horn ratio. As the scatterplot (Fig. 3A) shows, there is a significant positive correlation between fractional anisotropy of the corona radiata and ventricular size in chronic hydrocephalus (r = 0.89, r 2 = 0.036, p < 83). Mean diffusivity of the corona radiata was also computed for each study and plotted against frontal occipital horn ratio. The scatterplot (Fig. 3B) shows a small positive correlation between mean diffusivity and ventricular size in adult hydrocephalus (r = 69; r 2 = 6, which was not significant; p= 9). Correlation between ventricular size and axial and radial diffusivity The principal eigenvalue (axial diffusivity) of the corona radiata was calculated for each study and plotted against the frontal occipital horn ratio. The scatterplot (Fig. 3C) illustrates a strong positive correlation between axial diffusivity and the ventricular size in chronic hydrocephalus (r = 69, r 2 = 6, p = 0.03). The radial diffusivity of the corona radiata was calculated for each study and plotted against the frontal occipital horn ratio. The scatterplot (Fig. 3D) shows no significant correlation between ventricular volume and radial diffusivity measurements within the corona radiata in adult hydrocephalus (r = 0, r 2 = 0.022, p = 0.03). Correlation between diffusivity measurements of the corona radiata and patient age To determine the correlation between diffusivity metrics and patient age in this study, plotted patient age against the fractional anisotropy, mean diffusivity, and axial and radial diffusivity values of the corona radiata. There was little correlation between these variables in adults with chronic hydrocephalus: age and corona radiata fractional anisotropy (r = 39, r 2 = 0.057, p = 73), age and corona radiata mean diffusivity (r = 0.0, r 2 = 0.02, p = 0.0), age and corona radiata radial diffusivity (r = 0.028, r 2 < 0.00, p = 0.97), and age and corona radiata axial diffusivity (r = 90, r 2 = 0.084, p = 90). These findings are summarized in Table 2. Model Summary B SE B t F df p Adjusted r 2 r 2 Δ F Δ p Control group Frontal-occipital horn ratio fractional anisotropy (2,45) 8d Frontal-occipital horn ratio mean diffusivity (2,45) 0.043d Frontal-occipital horn ratio axial diffusivity (2,45) 0.003d Frontal-occipital horn ratio radial diffusivity (2,45) 0.085d Acute group Frontal-occipital horn ratio fractional anisotropy (2,) 0.00d Frontal-occipital horn ratio mean diffusivity (2,) 0.043d Frontal-occipital horn ratio axial diffusivity (2,) 0.054d Frontal-occipital horn ratio radial diffusivity (2,) 0.002d Chronic group Frontal-occipital horn ratio fractional anisotropy (2,3) 69d Frontal-occipital horn ratio mean diffusivity (2,3) 29d Frontal-occipital horn ratio axial diffusivity (2,3) 0.045d Frontal-occipital horn ratio radial diffusivity (2,3) 4d Note Independent variable is frontal occipital horn ratio and control variable is age. Arrow denotes independent variable to dependent variable. B = regression weight, SEB = standard error, df = degrees of freedom, r2 Δ = difference in r2 for model, F Δ = F change value. AJR:203, July

7 Cauley and Cataltepe 0 5 y = x r 2 = Age (y) Fig. 5 Scatterplot shows ventricular size as function of age in control subjects. Fractional anisotropy values of corona radiata are compared with ventricular size estimated with frontal occipital horn ratio with linear regression (coefficient of determination [r 2 ] = 0.069, p = 0.005). TABLE 4: Regression Model Fit Comparisons Paths Group (r values) Control Chronic Acute Diffusion-Tensor Characteristics of Corona Radiata in Control Subjects Correlation between ventricular size and fractional anisotropy and mean diffusivity Fractional anisotropy and mean diffusivity of the corona radiata were measured and plotted against the frontal occipital horn ratio in 48 consecutive adult subjects. As the scatterplot (Fig. 4A) shows, there was no significant correlation between ventricular size and fractional anisotropy of the adjacent corona radiata in the control group (fractional anisotropy, r = 0.84, r 2 = 0.034, p = 00. There was a significant positive correlation (Fig. 4B) between ventricular size and mean diffusivity in the control group, (r = 82, r 2 = 0.079, p = 0.047). Correlation between ventricular size and axial and radial diffusivity Axial and radial diffusivity of the corona radiata were measured for each case in the control group and plotted against the frontal occipital horn ratio. The scatterplots show a significant positive correlation between axial diffusivity of the corona radiata and ventricular size (r = 46, r 2 = 0.99, p < 0.00) (Fig. 4C) and no significant correlation between radial diffusivity of the corona radiata and ventricular size, (r = 0.07, r 2 = 0.005, p = 5) (Fig. 4D). The axial diffusivity increases more significantly with a frontal occipital horn ratio greater than 5, and a curvilinear polynomial regression offers a higher correlation with r 2 = 39. Correlation between ventricular size, diffusivity measurements of the corona radiata, and age of control subjects Because patient age represented a variable in this study, we performed regression analysis to determine the correlation between patient age and ventricular size as well as between patient age and diffusivity measurements of the adjacent periventricular WM tracts of the corona radiata (Fig. 5). We found no significant correlation between patient age and ventricular size (r = 62, r 2 = 0.069, p = 0.066) but a significant correlation with radial diffusivity (r = 22, r 2 = 0.04, p = 0.026) and no significant correlation between patient age and fractional anisotropy, mean diffusivity, or axial diffusivity of the corona radiata in this study population (Table 2). Correlation between corona radiata diffusivity measurements in patients with acute hydrocephalus, patients with long-standing hydrocephalus, and control subjects Figure 6 shows a scatterplot of data in acute hydrocephalus (as in Fig. 2) and chronic hydrocephalus (as in Fig. 3) together with the data from nonhydrocephalus patients (as in Fig. Control vs Chronic Fisher z Test Control vs Acute Chronic vs Acute Frontal occipital horn ratio fractional anisotropy Frontal occipital horn ratio mean diffusivity Frontal occipital horn ratio axial diffusivity Frontal occipital horn ratio radial diffusivity Sample size Note Arrow denotes independent variable to dependent variable. Axial Diffusivity 2.2 y =.98x + 97 r 2 = y =.8843x + 35 r 2 = 69 y =.4305x r 2 = Acute Control Linear (chronic) Chronic Linear (acute) Linear (control) Fig. 6 Composite scatterplot of axial diffusivity in white matter of corona radiata in acute hydrocephalus, long-standing (chronic) hydrocephalus, and control subjects. Slope of regression line is similar for each cohort. 3), with regression lines. In acute hydrocephalus, the slope of the regression line is.43; in chronic hydrocephalus, the slope is.88; and in healthy subjects, the slope is.9. Results of Multifactorial Regression Analysis Table 3 presents a summary of the series of hierarchic regression analysis conducted for each group. Controlling for the effect of age, the prediction models for each group were statistically significant in some specific paths. In particular, the effect of frontal occipital horn ratio on axial diffusivity (λ // ) was statistically significant for all the groups: control group (B = 73, p = 0.00; r 2 Δ = 20, F Δ = 2.744, p = 0.00), acute hydrocephalus (B = 45, p = 0.08; r 2 Δ = 0.847, F Δ = 95, p = 0.00), and long-standing hydrocephalus (B = 83, p = 0.027; r 2 Δ = 0.847, F Δ = 95, p = 0.027). For the acute group, the effect of the frontal occipital horn ratio on fractional an- Lateral Ventricle Corona Radiata Lateral Ventricle Corona Radiata Fig. 7. Schematic shows model proposed to explain correlation of ventricular size and axial diffusivity of white matter fibers of corona radiata. Size of lateral ventricle is proportional to packing density of white matter fibers. 76 AJR:203, July 204

8 Imaging Adult Hydrocephalus TABLE 5: Differences in Regression Weights Across Groups Control Chronic Acute Normal vs Chronic Normal vs Acute Paths isotropy and radial diffusivity (λ ) were also statistically significant. Results of the subsequent Fisher z test to compare the fit of the regression models are summarized in Table 4. On the basis of the findings, it can be concluded that the regression model applies to the control group in comparison with both the long-standing and acute groups (with z values of p > 0.05) B SE B SE B SE SE (B-difference) z SE (B-difference) z Age Frontal occipital horn ratio fractional anisotropy Frontal occipital horn ratio mean diffusivity Frontal occipital horn ratio axial diffusivity Frontal occipital horn ratio radial diffusivity Note Arrow denotes independent variable to dependent variable. B = regression weight, SE = standard error, SEB-difference = standard error of regression weights between groups. TABLE 6: Correlation Matrix: Zero-Order Correlations in some select effects. More specifically, the groups are comparable for the effects of frontal occipital horn ratio fractional anisotropy and as an independent variable on radial diffusivity (λ ). Comparability of the individual contributions of the predictors was further assessed (Table 5). Findings generally indicated that the regression weights did not significantly Variables differ across the groups except for the following: frontal occipital horn ratio as an independent variable on fractional anisotropy where the regression weights between the control group and the acute group significantly differed (z = 2.687, p < 0.05) and in the case of frontal occipital horn ratio as an independent variable on radial diffusivity (λ ) (z = 3.229, p < 0.0). Semipartial Correlation With Control group (n = 48). Frontal occipital horn ratio 3 2. Fractional anisotropy a 3. Mean diffusivity 07 a a 4. Axial diffusivity 66 b 44 b 98 b Radial diffusivity b b a 6. Age 80 a 9 a a 3 Acute group (n = 4). Frontal occipital horn ratio 2. Fractional anisotropy 84 b Mean diffusivity Axial diffusivity 38 b Radial diffusivity 68 b 0.90 b 52 b Age Chronic group (n = 6). Frontal occipital horn ratio 2. Fractional anisotropy Mean diffusivity b Axial diffusivity 07 a b Radial diffusivity b b 27 b Age 07 a a p < p < 0.0. AJR:203, July

9 Cauley and Cataltepe For the main findings (Table 6), frontal occipital horn ratio had positive correlations with mean diffusivity (r = 07, p < 0.05) and λ // (r = 66, p < 0.0) in the control group. Furthermore, the semipartial correlations for both mean diffusivity and λ // were also statistically significant. This implies that both of these variables had significant contributions to changes in the frontal occipital horn ratio even after isolating the overlapping effects of the other variables. The semipartial correlation between frontal occipital horn ratio and age was also significant (partial r = 90. p < 0.05). Moreover, the squared value of the semipartial correlation coefficient represents the percentage of variance in the frontal occipital horn ratio that could be uniquely due to that specific predictor. In the cases of the chronic and the acute groups, the frontal occipital horn ratio had a significant positive correlation with λ // and age. It was noted, however, that the respective semipartial contributions were not statistically significant. Discussion Previous studies have shown that diffusion-tensor characteristics of periventricular WM are altered in hydrocephalus [, 2, 0, ]. We investigated this observation to determine the nature of the correlation between ventricular size and DTI metrics of the corona radiata. We found a direct correlation between ventricular size and changes in DTI metrics in hydrocephalus. Furthermore, we showed that the linear regression of DTI metrics against ventricular size differs between cases of hydrocephalus classed as either acute or long-standing. Finally, we established a correlation between ventricular size and DTI metrics in healthy subjects. Others have reported increased fractional anisotropy and increased axial diffusivity values in the corona radiata in pediatric patients with obstructive hydrocephalus [] and in adult patients with idiopathic normalpressure hydrocephalus [0 2]. We wanted to investigate these qualitative observations to determine whether there may be some underlying rules regarding the relationship between hydrocephalus and diffusion changes in the adjacent WM. Because hydrocephalus presents with varying degrees of severity according to the degree of ventricular enlargement, we propose that diffusion-tensor changes in the WM of the ascending fibers of the corona radiata are proportional to ventricular size. Identification of such a relationship may aid in distinguishing the various types of hydrocephalus and may aid in understanding the diffusion changes seen in the periventricular WM of hydrocephalic patients. Such a correlation model may serve to identify cases in which diffusion-tensor metrics deviate significantly from the values predicted by the model, with the potential to identify cases in which WM damage is irreversible and invasive procedures, such as shunting, are less likely to be of benefit. Our data identify a linear relationship between ventricular size and axial diffusivity of the WM fibers of the adjacent corona radiata that is significantly correlated with ventricular size in acute hydrocephalus, chronic hydrocephalus, and normal control subjects. The slope of this regression line is similar in all three patient populations. In acute hydrocephalus, the radial diffusion significantly decreases and fractional anisotropy significantly increases with ventricular size. There was no significant correlation between radial diffusion or fractional anisotropy and ventricular size in chronic hydrocephalus or in normal control subjects. Because brain atrophy and increased ventricular size are correlated with aging, we investigated the contribution of age to our observations. The coefficient of determination (r 2 ) represents the percentage of the variation in the dependent variable that is explained by the relationship depicted by the regression line. Our analysis using simple linear regression shows that 22% of the axial diffusion of the corona radiata in healthy subjects can be attributed to ventricular size; no significant correlation with age could be found for fractional anisotropy or axial or radial diffusivity of the corona radiata WM fibers in our cohort. We did not find a significant decline in the anisotropy of the corona radiata as a function of patient age. Studies designed to reveal the effects of aging on anisotropy of the fibers of the corona radiata, corticospinal tract, or internal capsule show that decreases in anisotropy as a function of age are small [0, 3 7] or are not significant [8, 9]. Some studies have shown left or right lateralization of the degree of change as a function of age [5, 20]. Differences in study design in terms of cohort size and subject age distribution; inclusion or exclusion criteria; and postprocessing variables, such as measurement technique, ROI size and placement, and averaging of left- and right-sided values, likely contribute to our finding of no statistically significant correlation between age and WM anisotropy of the corona radiata. WM anisotropy is generally believed to decline with age [2]. Our observations propose that a second variable ventricular size is a strong determinant of axial diffusivity, and increased ventricular size results in increased axial diffusivity. Because brain volume loss and increased ventricular size are seen with aging [2] and appear to be stronger determinants of anisotropy of the corona radiata than aging from our dataset, data from a broad range of ages might be best characterized by a curvilinear correlation, with increased axial diffusivity with increasing age. A recent study of axial diffusivity values of the corticospinal tract as a function of age in healthy volunteers showed a curvilinear correlation with age, with lowest axial diffusivity values seen at approximately age 50 years, with increasing axial diffusivity values seen in older patients [6, 2]. The curved correlation reported in this study may be explained by increasing ventricular size with age. Fiber compression resulting in greater alignment of fibers with higher fiber-packing density is the likely explanation for increased axial diffusivity with increased ventricular size [, 0], although this explanation remains theoretic [8, 22, 23]. Fiber-packing density is related to the brain parenchymal thickness rather than to increased intracranial pressure. Viewed in this way, the correlation with ventricular size is seen not only in pathologic states of hydrocephalus but also in healthy subjects, and the slope of the regression line is nearly identical for control subjects and for patients with hydrocephalus (Figs. 5 and 7). The correlation persists for individuals with ventricular size that is smaller than average, further supporting the correlation between axial diffusivity of the corona radiata and absolute ventricular size rather than ventricular enlargement (Figs. 3 and 5). Study Limitations Hydrocephalus patients are a heterogeneous group in terms of cause, duration, and severity of the condition as well as comorbidities. We have arbitrarily created two classes of hydrocephalus patients acute and longstanding in an attempt identify differences between these populations in terms of periventricular WM characteristics. Larger cohort studies may increase the statistical power of the claims and may enable characterization of smaller populations, such as the normal-pressure hydrocephalus group. The study cohorts represent consecutive cases and are not age-matched. The average age of chron- 78 AJR:203, July 204

10 Imaging Adult Hydrocephalus ic hydrocephalus patients (54.3 years) was significantly older than that of acute hydrocephalus patients (42.5 years) or control subjects (44.7 years). Age was therefore correlated as a separate variable. Future studies should attempt to correlate treatment outcomes with diffusion-tensor metrics before and after intervention. Our healthy subjects were imaged for a clinical indication and therefore were not truly healthy; however, our inclusion criteria were created to minimize the effects of this type of study limitation and to permit inclusion of a large number of control subjects. Conclusion Axial diffusivity in the ascending WM fibers of the corona radiata is linearly correlated with ventricular size in healthy subjects and patients with hydrocephalus. Radial diffusivity decreases with ventricular size in acute hydrocephalus but not in chronic hydrocephalus. Diffusion measurements may aid in characterizing hydrocephalus and have implications for optimizing treatment strategies. Acknowledgment We thank Ron Fisher for assistance with statistical analysis. References. Assaf Y, Ben-Sira L, Constantini S, Chang LC, Beni- Adani L. Diffusion tensor imaging in hydrocephalus: initial experience. AJNR 2006; 27: Hakim S, Adams RD. The special clinical problem of symptomatic hydrocephalus with normal cerebrospinal fluid pressure: observations on cerebrospinal fluid hydrodynamics. J Neurol Sci 965; 2: O Hayon BB, Drake JM, Ossip MG, Tuli S, Clarke M. Frontal and occipital horn ratio: a linear estimate of ventricular size for multiple imaging modalities in pediatric hydrocephalus. Pediatr Neurosurg 998; 29: Softpedia website. MidINRIA.9.0 free download. Download html. Accessed March 26, Basser PJ, Pierpaoli C. A simplified method to measure the diffusion tensor from seven MR images. Magn Reson Med 998; 39: Pierpaoli C, Basser PJ. Toward a quantitative assessment of diffusion anisotropy. Magn Reson Med 996; 36: Basser PJ, Mattiello J, LeBihan D. Estimation of the effective self-diffusion tensor from the NMR spin echo. J Magn Reson B 994; 03: Pierpaoli C, Jezzard P, Basser PJ, Barnett A, Di Chiro G. Diffusion tensor MR imaging of the human brain. Radiology 996; 20: Wakana S, Jiang H, Nagae-Poetscher LM, van Zijl PC, Mori S. Fiber tract-based atlas of human white matter anatomy. Radiology 2004; 230: Hattingen E, Jurcoane A, Melber J, et al. Diffusion tensor imaging in patients with adult chronic idiopathic hydrocephalus. Neurosurgery 200; 66: Hattori T, Ito K, Aoki S, et al. White matter alteration in idiopathic normal pressure hydrocephalus: tractbased spatial statistics study. AJNR 202; 33: Kanno S, Abe N, Saito M, et al. White matter involvement in idiopathic normal pressure hydrocephalus: a voxel-based diffusion tensor imaging study. J Neurol 20; 258: Burzynska AZ, Preuschhof C, Backman L, et al. Age-related differences in white matter microstructure: region-specific patterns of diffusivity. Neuroimage 200; 49: Hugenschmidt CE, Peiffer AM, Kraft RA, et al. Relating imaging indices of white matter integrity and volume in healthy older adults. Cereb Cortex 2008; 8: Kawaguchi H, Obata T, Ota M, et al. Regional heterogeneity and age-related change in sub-regions of internal capsule evaluated by diffusion tensor imaging. Brain Res 200; 354: Sala S, Agosta F, Pagani E, Copetti M, Comi G, Filippi M. Microstructural changes and atrophy in brain white matter tracts with aging. Neurobiol Aging 202; 33: Salami A, Eriksson J, Nilsson LG, Nyberg L. Age-related white matter microstructural differences partly mediate age-related decline in processing speed but not cognition. Biochim.Biophys. Acta 202; 822: Bhagat YA, Beaulieu C. Diffusion anisotropy in subcortical white matter and cortical gray matter: changes with aging and the role of CSF-suppression. J Magn Reson Imaging 2004; 20: Madden DJ, Whiting WL, Huettel SA, White LE, MacFall JR, Provenzale JM. Diffusion tensor imaging of adult age differences in cerebral white matter: relation to response time. Neuroimage 2004; 2: Ardekani S, Kumar A, Bartzokis G, Sinha U. Exploratory voxel-based analysis of diffusion indices and hemispheric asymmetry in normal aging. Magn Reson Imaging 2007; 25: Gunning-Dixon FM, Brickman AM, Cheng JC, Alexopoulos GS. Aging of cerebral white matter: a review of MRI findings. Int J Geriatr Psychiatry 2009; 24: Beaulieu C. The basis of anisotropic water diffusion in the nervous system: a technical review. NMR Biomed 2002; 5: Le Bihan D. Looking into the functional architecture of the brain with diffusion MRI. Nat Rev Neurosci 2003; 4: Relkin N, Marmarou A, Klinge P, Bergsneider M, Black PM. Diagnosing idiopathic normal-pressure hydrocephalus. Neurosurgery 2005; 57(3 suppl):s4 S6 AJR:203, July