Brain Perfusion Imaging Under Acetazolamide Challenge for Detection of Impaired Cerebrovascular Reserve Capacity: Positive Findings with ¹⁵O-Water PET in Patients with Negative ^99mTc-HMPAO SPECT Findings

Eligibility Criteria

Our database was searched according to the following criteria: SPECT with ^99mTc-HMPAO under acetazolamide challenge had been performed to assess CVRC for presurgical evaluation of symptomatic MMV or ACVD, the SPECT findings had been interpreted as “no evidence of impaired CVRC,” the negative SPECT findings had been suspected to be false-negative by the referring neurosurgeon on the basis of the patient's symptoms, and assessment of CVRC had been repeated using ¹⁵O-water PET within 4 wk.

This retrospective study was approved by the Ethics Committee of Charité–Universitätsmedizin Berlin (reference number EA4/142/16), and the requirement to obtain informed consent was waived.

^99mTc-HMPAO SPECT

Resting brain perfusion SPECT was performed 40 min after intravenous administration of 510.3 ± 62.2 MBq of ^99mTc-HMPAO. On the next day or the day afterward, perfusion SPECT was repeated 5–15 min after infusion (2.5 mL/min) of 1,000 mg of acetazolamide in 50 mL of 0.9% NaCl (513.3 ± 62.4 MBq; paired t test, P = 0.708). A GE Healthcare Millennium VG Hawkeye or a Siemens MultiSPECT 3 camera was used for SPECT (same camera for resting and acetazolamide scan in each patient). Images were reconstructed using filtered backprojection with the same Butterworth filter in all cases (cutoff, 0.5 cycles/cm; 10th order; Chang attenuation correction with μ = 0.12 cm⁻¹; no scatter correction). Spatial resolution in the reconstructed SPECT images was about 13 mm in full width at half maximum.

¹⁵O-Water PET

Brain perfusion PET was performed with 700 MBq of ¹⁵O-water and a Philips Gemini TF16 PET/CT system. The time of arrival of ¹⁵O-water in the brain was determined from the whole-brain time–activity curve. A static ¹⁵O-water uptake image was obtained by iterative reconstruction of 40 s of list-mode data starting at this time point (3-dimensional line-of-response algorithm of the system software with default parameter settings, CT-based attenuation correction). Infusion of acetazolamide was initiated immediately after the resting perfusion PET scan using the same dose and infusion protocol as for the SPECT scan. ¹⁵O-water PET was repeated 5–15 min after acetazolamide infusion. Spatial resolution in the reconstructed PET images was about 7 mm in full width at half maximum.

Parametric Maps of Relative CVRC (rCVRC)

rCVRC was estimated voxelwise to generate a 3-dimensional rCVRC map (Fig. 1). This was done separately for SPECT and PET using an automated processing pipeline comprising the following steps. First, the SPECT or PET image is smoothed with an isotropic 3-dimensional gaussian kernel of 8 mm (SPECT) or 6 mm (PET) in full width at half maximum. Second, the Statistical Parametric Mapping (SPM8) “coregister” tool (Wellcome Trust Centre for Neuroimaging) is used to estimate the rigid-body transformation that maps the acetazolamide image to the resting image. Third, the SPM8 “normalize” tool is used to map the resting and the coregistered acetazolamide image into SPM’s template space. Fourth, images are smoothed with an isotropic gaussian kernel of 20 mm in full width at half maximum. Finally, the rCVRC map is computed voxel by voxel asEq. 1where swA and swR denote the smoothed, anatomically standardized uptake image under acetazolamide or resting conditions, respectively. The computation of rCVRC is restricted to a binary cerebrum mask that was obtained by thresholding the SPM8 gray matter probability map at 0.3.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Standardized display to support visual interpretation of ^99mTc-HMPAO SPECT (top) and ¹⁵O-water PET (bottom) with respect to impairment of rCVRC.

The scale factor sf in Equation 1 accounts for (small) variations in the administered tracer dose between resting and acetazolamide conditions. It is determined such that the average rCVRC over a reference region (defined below) is 30% (2), that is,Eq. 2where mean_Ref is the mean value over the reference region. The reference region is intended to include all “normal” voxels within the cerebrum mask, that is, voxels with normal perfusion both in the resting and in the acetazolamide scans. Normal voxels were determined independently for resting and acetazolamide scans using the following iterative 2-step approach and starting with all voxels in the cerebrum mask. First, all voxels with an intensity below the fifth percentile were removed. Second, the mean intensity of the remaining voxels was compared with the mean intensity of the normal voxels in the preceding iteration. The 2 steps were repeated until the mean intensity of normal voxels changed less than 1%. The iterative procedure worked properly according to visual inspection in all cases (contour delineating normal voxels overlaid to transversal images). The reference region was obtained as the intersection of normal voxels from the resting scan and normal voxels from the acetazolamide scan.

Visual Interpretation

Visual interpretation of SPECT and PET was supported by a PDF document (created by the automatic processing pipeline) with standardized side-by-side display of anatomically normalized resting and acetazolamide images together with the rCVRC map (Fig. 1). Visual interpretation with respect to presence or absence of CVRC impairment was performed independently by 2 experienced readers masked to clinical information.

Quantitative Analysis of CVRC Defects

CVRC defects were defined by thresholding the rCVRC map at 15% (i.e., 50% reduction from the normal rCVRC of 30%). The volume of the CVRC defect and the mean rCVRC within the defect were computed.