Between October 2021 and June 2023, 24 consecutive patients with acute CSVT were admitted to our institution. Thirteen of these patients, who underwent [15O]water PET for examination of individual regional CBF, were eligible for the present retrospective analysis. However, one patient was previously treated for a frontal glioblastoma. Because of residual changes in the frontal lobe, this patient was excluded from further analysis, leaving a total of twelve patients being included. [15O]water PET was not performed in the other eleven patients for the following reasons: five were in a bad clinical condition requiring prolonged intensive care or had a severe prognosis due to associated metastatic cancer. Three patients refused to undergo [15O]water PET as an additional imaging procedure, one patient was pregnant and one patient had to be isolated due to acute SARS-CoV2 infection. Ten healthy controls (HC) from an ongoing prospective study scanned under comparable conditions served as a control group. Scans were acquired in duplicate to account for the relatively high noise level of [15O]water PET.

After diagnosis of CSVT by contrast-enhanced MRI or CT venography, all patients received immediate therapeutic anticoagulation with unfractionated intravenous or low molecular weight heparin when admitted to our stroke unit or neurological intensive care unit. Parenteral anticoagulation was replaced by oral anticoagulation with vitamin K antagonists (INR 2.0–3.0) or off-label treatment with dabigatran (2 × 110/150 mg per day) at discharge from hospital and continued for at least six months. Follow-up brain MRI and neurological examination were performed six months later in 6/12 patients. The study has been approved by the institutional review board and the need for written informed consent to this retrospective analysis was waived (vote no. 23-1549-S1-retro).

Two 4-min list mode scans under resting conditions at ambient light and noise were acquired in patients and healthy controls, starting with intravenous injection of 300 ± 14 MBq [15O]water on a fully-digital PET/CT (Philips, Vereos). PET datasets were reconstructed employing low-dose CT for attenuation correction and the vendor-specific, line-of-response time-of-flight ordered-subsets 3-dimensional iterative reconstruction algorithm using spherically symmetric basis functions (so-called blob ordered-subset time-of-flight reconstruction; number of iterations, 5; number of subsets, 11; 2 mm Gaussian post filtering; point spread function off; resulting voxel size, 1.0 mm3), yielding an isotropic voxel size of 1 × 1 × 1 mm3. Static images over one minute after tracer arrival in the brain (i.e., 15–75 s) were calculated and normalized to the cerebellum to gain maps of relative CBF (rCBF). These maps were used for further voxel- and region-of-interest (ROI)-based analyses (see below). We mirrored image datasets of patients with left-sided CSVT to harmonize the affected hemisphere (i.e., right).

All preprocessing steps were implemented with an in-house pipeline in MATLAB (The MathWorks, Inc.) and Statistical Parametric Mapping (SPM) 12 software (www.fil.ion.ac.uk/spm). For each individual, the second PET scan was realigned to the first PET scan to correct for possible inter-scan misalignment. Realigned scans were spatially normalized to an in-house PET template in MNI space.

Voxel-wise comparisons between groups were performed in SPM using full-factorial design accounting for repeated measures. For this analysis, rCBF maps were smoothed with an isotropic Gaussian kernel of 10 mm full-width at half maximum. Voxels with false-discovery rate (FDR)-corrected p < 0.05 and a cluster extent of at least 0.1 mL (corresponding to 100 voxels) were considered statistically significant.

Mean rCBF in regions from the automatic anatomic labeling (AAL3) atlas [11] was calculated in each patient and healthy control. Selected AAL3-defined regions were grouped to create bilateral cortical ROI as follows: frontal cortex (superior, middle, and inferior frontal gyri), parietal cortex (superior and inferior parietal gyri), occipital cortex (superior, middle, and inferior occipital gyri), temporal cortex (superior, middle, and inferior temporal gyri), and mesial temporal lobe (hippocampus, parahippocampal gyrus and amygdala). In addition, we assessed mean rCBF in bilateral caudate nucleus, putamen, and thalamus (see Fig. 1). Since spatial normalization and ROI placement on caudate nuclei may be inaccurate in patients with enlarged ventricles, we carefully checked the position of the caudate nucleus ROI and adjusted their position in mediolateral (x) direction if felt necessary.

Anatomical definitions of the corresponding masks employed for region of interest analysis

Statistical analyses were performed with R software (version 4.1.0, http://www.R-project.org/). Group differences in mean rCBF of aforementioned ROI were tested with repeated-measures ANOVA followed by pairwise least square means comparison with Tukey multiplicity adjustment. For all comparisons, adjusted p < 0.05 was considered statistically significant.