Cerebral Blood Flow by Using Pulsed Arterial Spin-Labeling in Elderly Subjects with White Matter Hyperintensities

Subjects

We included 21 subjects (8 men, 13 women) from a single center within the Leukoaraiosis and Disability (LADIS) study.⁸ The current study was approved by both the local and the LADIS study institutional review boards, as well as by the LADIS study steering committee. All subjects gave written informed consent.

The LADIS study is a prospective longitudinal multicenter European study on the role of WMH as an independent predictor for transition to disability in nondemented elderly subjects.⁸ Inclusion criteria for the LADIS study were the following: age between 65 and 84 years, WMH on MR imaging of any degree according to the Fazekas scale,⁹ absence of disability, presence of a regularly contactable informant, and agreement to sign an informed consent. Exclusion criteria were the following: presence of severe illnesses, severe unrelated neurologic diseases, leukoencephalopathy of nonvascular origin, severe psychiatric diseases, inability to give an informed consent, and inability or refusal to undergo cerebral MR imaging.

All subjects underwent an MR imaging examination at baseline. In our center, follow-up MR imaging examinations were performed yearly, and the current report focuses on data obtained during the second or third year of follow-up.

On the basis of clinical findings at the time of undergoing MR imaging, the included subjects for the current study were classified as having normal cognition or cognitive impairment.^10–15 Concurrent medications and drugs possibly influencing CBF were also taken into account.

MR Imaging Protocol

MR imaging examinations were performed at 1.5T (Sonata; Siemens, Erlangen, Germany) and included axial FLAIR images (TE = 84 ms; TR = 9000 ms; NEX = 2; TI = 2200 ms; FOV = 250 mm; section thickness = 5 mm; number of sections = 24; acquisition matrix = 256 × 192, interpolated to image matrix = 512 × 384) and a high-resolution 3D T1-weighted inversion recovery sequence (TE = 5.17 ms; TR = 2700 ms; TI = 950 ms; flip angle = 8 °; NEX = 1; FOV = 250 mm; section thickness = 1 mm; number of sections = 160; acquisition matrix = 256 × 176 [69% FOV phase], interpolated to image matrix = 512 × 352). In addition, to obtain relative CBF (rCBF) images, we chose a quantitative imaging of perfusion by using a single-subtraction second version, with a thin-section TI periodic saturation (Q2TIPS) PASL sequence with proximal inversion and control for off-resonance effects.¹⁶ To convert the signal intensity of the rCBF images to absolute values of CBF, we performed a single-shot EPI sequence of the fully relaxed brain tissue.

Q2TIPS Tailored to an Elderly Population

ASL techniques rely on subtracting control images (acquired with blood and tissue water in identical magnetization states) from spin-labeled images (acquired with blood and tissue water in different magnetization states). The Q2TIPS variant of PASL enables the acquisition of images in multiple sections and aims to control for 2 major systematic errors of ASL: the variable transit time from the distal edge of the labeled region to the image sections, and the contamination by intravascular signal intensity from labeled blood that flows through the image sections by applying thin-section periodic saturation pulses at the distal end of the labeled region.¹⁶

We used a 10-cm labeling region, a TE = 15 ms, a TI₁ = 700 ms (time between the inversion pulse and the beginning of periodic saturation pulses), a TI₁ stop time (TI_1s) = 1600 ms (time between the inversion pulse and the end of periodic saturation pulses), a TI₂ = 1800 ms (time between the inversion pulse and acquisition of the proximal image), and a TR = 2500 ms. By using a TI₂ = 1800 ms, one can obtain an appropriate brain tissue signal intensity in elderly subjects, almost without intravascular signal intensity.¹⁷ To further improve suppression of intravascular signal intensity, we additionally used a bipolar crusher gradient. The sequence was repeated 200 times, alternating between acquisition of 100 labeled and 100 control images (total imaging time = 8.33 minutes), and prospective motion correction¹⁸ was applied before subtracting each labeled-control pair. After subtraction, the difference images were averaged to obtain rCBF images in 6 sections (FOV = 224 mm, section thickness = 6 mm, intersection gap = 1.5 mm, matrix = 64 × 64, ascending section acquisition order).

Quantification of CBF

Quantification of CBF by means of ASL is based on the difference in longitudinal magnetization (ΔM) between labeled and control images. For Q2TIPS, ΔM at time t = TI₂ is related to CBF by the following formula^16,19: 1) TI₁ and TI₂ are sequence parameters. T1_b is the T1 value of arterial blood, α is the degree of spin inversion, m⁰_a is the arterial blood longitudinal magnetization at equilibrium, t_L is the time width of the labeling, t_T is the transit time from the distal edge of the labeled region to the image sections, and f corresponds to absolute values of CBF. We assumed that TI₁ should be <t_L and that TI₂ should be >TI₁ + t_T to obtain valid CBF measurements. In addition, we assumed T1_b = 1400 ms and α = 0.97.^16,19

To account for differences in proton density between blood and brain tissue, we assumed an average value (between gray and white matter) of 0.90 for the blood/brain partition coefficient (λ), which is defined by the following formula^19,20: 2) M⁰ corresponds to the brain tissue longitudinal magnetization at equilibrium¹⁹ and is related to the magnitude of signal intensity at equilibrium (S⁰) by the scanner gain G: 3) S⁰ was determined by using a single-shot EPI sequence of the fully relaxed brain tissue. This single-shot EPI sequence has an EPI readout similar to that of Q2TIPS but does not have any presaturation, labeling, or periodic saturation pulses. In addition, a dead time of 10 seconds was inserted between adjustment pulses of the scanner and the acquisition of images.

Finally, the difference in signal intensity (ΔS) between labeled and control images (signal intensity of rCBF images) is also related to ΔM by the scanner gain G: 4)

Combining equations 1, 3, and 4 and additionally accounting for T2* effects of blood and brain tissue, we calculated a scaling factor C to determine absolute CBF measurements from ΔS and S⁰: 5)

Values for T2* of blood (40 ms) and brain tissue (100 ms, gray and white matter average) were taken from the literature.²¹ Using these values and the values of TE, TI₁, TI₂, T1_b, α, and λ listed above, C was calculated to be 11.5 · 10³ mL/100 mL/min for the proximal section.

Image Analysis

A single reader, with more than 5 years of experience, visually rated the degree of WMH on axial FLAIR images. On the basis of the Fazekas scale,⁹ WMH were classified into the following categories: punctiform (score 1), beginning confluent (score 2), and diffuse confluent (score 3).

For all the included cases, the rCBF and S⁰ image sections were uniformly positioned, parallel to the anterior/posterior commissure line, to include the deep gray matter and most of the cerebral white matter. To analyze rCBF and S⁰ images, we used a statistical parametric mapping program (SPM5; Wellcome Department of Cognitive Neurosciences, London, UK). For each subject, rCBF and S⁰ images were coregistered²² to the T1-weighted images by using a stepwise coregistration involving first the coregistration of mean EPI images, obtained after averaging all labeled and control images, to the T1-weighted images. Additionally, a segmentation algorithm²³ combining anatomic information and signal intensity was applied to the T1-weighted images to obtain probabilistic gray and white matter maps. After normalization^24,25 to the standard T1-weighted template, gray and white matter probability maps were converted to binary masks.

Because most of WMH are isointense with gray matter on T1-weighted images, the corresponding voxels have higher probability values on the gray matter maps and lower probability values on the white matter maps than voxels representing normal-appearing white matter. Therefore, to reduce misclassification of white matter lesions as gray matter after brain segmentation, we used a lower threshold to obtain white matter binary masks than to obtain gray matter binary masks (ie, voxels with values >40% on the gray matter probability maps were assigned a value 1 on the gray matter binary masks, and voxels with values >20% on the white matter probability maps were assigned a value 1 on the white matter binary masks). The use of a 40% threshold to obtain gray matter binary masks was also planned to minimize partial volume effects of the CSF occurring, particularly in cases with cortical atrophy. On the obtained binary masks, we still created regions of interest either to correctly classify voxels representing lesions (eg, voxels representing white matter lesions misclassified as gray matter) or to exclude nonbrain voxels.

Because both white matter and deep gray matter hyperintensities are considered to represent subcortical ischemic small vessel disease, we decided to combine voxels classified as deep gray matter and voxels classified as white matter into common subcortical binary masks.

The cortical gray matter and subcortical binary masks were multiplied by the normalized rCBF images, as well as global binary masks obtained after combining all gray matter and white matter. Global binary masks were also multiplied by normalized S⁰ images. Then, the average intensity of the product images was taken, and on the basis of equation 5, values of cortical gray matter, subcortical (including white matter and deep gray matter), and global CBF were calculated. For all CBF calculations, we used a global average (combining all gray matter and white matter) S⁰ intensity measurement.

Statistical Analysis

Statistical analysis was performed by means of the Statistical Package for the Social Sciences 11.0 for Windows (SPSS, Chicago, Ill). Subjects with Fazekas score 3 were compared with those with Fazekas score 1 or 2, because there were few subjects with Fazekas score 1 and the sample size was limited. Given that the Shapiro-Wilk W test demonstrated normality in the distribution of data, we used the independent-samples Student t test to compare continuous variables, such as CBF measurements. To compare proportions, we used the Fisher exact test. The existence of a sex imbalance between the 2 groups analyzed led us to carry out a multiple linear regression analysis, to account for the effect of sex on CBF. Statistical significance was considered when P values were < .05.