This study was approved in 2023 by the local Ethics Committee. Patient data was pseudonymized. Due to local data privacy laws, data cannot be exported from the cloud-based cybersecure operating environment Acamedic [14], which has been designed for storing, processing, and analyzing research data of patients. We did not receive any funding for this study, and we do not report any conflicts of interest. The study was conducted in accordance with the Helsinki Declaration [15].

All adult patients who underwent GTR of supratentorial IM at the study Hospital between 2000 and 2020 were retrospectively identified. Patient identification and study material collection were conducted using electronic patient records stored in various medical software systems, such as Uranus (CGI, Helsinki, Finland), Opera (GE Healthcare, Chicago, Illinois, United States), RADU (L-Force, Helsinki, Finland), picture archiving and communication system (PACS) [16] and Qpati (Tietoevry, Espoo, Finland). First, all patients with the International Classification of Diseases 10 (ICD-10) diagnosis [17] code D32 (benign neoplasm of cerebral meninges) were identified through the Uranus database. Next, these patients’ digital MRI images were fetched from the PACS and transferred into the Acamedic platform.

IM location was initially determined based on surgical reports, and confirmed by reviewing the MRI images. IM locations were divided into five groups: convexity, falx, parasagittal, skull base IMs, and others [18]. The group “others” included intraventricular and tentorial IMs. Anterior clinoid, olfactory groove, planum sphenoidale, and sphenoid wing IMs were considered as skull base IMs. IM laterality was classified as left, right or bilateral, with the latter indicating that the IM extended across the midline. Based on the neuropathology reports, IMs were classified into 15 histological subtypes according to WHO [2]. The location of PTBE was classified based on the brain lobule most affected, categorizing it into frontal, temporal, parietal, and occipital regions.

Patients were required to meet the following inclusion criteria: (1) GTR of a histologically confirmed IM, (2) first intracranial surgery, (3) pre- and postoperative MRI available (minimum of a 1.5T scanner, contrast-enhanced preoperative images, FLAIR sequences), (4) no other intracranial tumors, (5) no radiotherapy administered at any point (radiation may lead to radiation-induced PTBE-like gliosis [19]), (6) age 18 or older, (7) presence of preoperative PTBE in last MRI before surgery, (8) surgery within one year of preoperative MRI, and (9) MRI follow-up time of at least one year. Extent of resection was estimated based on surgical reports and an in-hospital head CT scan. Furthermore, if any residual or recurring IMs or other intracranial tumors were identified during the follow-up, the follow-up was discontinued.

IM and PTBE volume measurements were based on MRI segmentations [20]. Segmentations were done manually on MR images using the open-source software 3D Slicer (version 5.4.0) [21]. Examples of segmentations are depicted in Fig. 1. After segmentations, 3D Slicer was used to create 3D objects of the IM and surrounding edema (Fig. 1D). Volume and surface-area calculations were done automatically by 3D slicer based on segmented voxels. Edema index (EI) was calculated manually as \(\:EI\:=\:\frac\) [22].

According to the hospital’s follow-up protocol, patients with gross-totally resected WHO grade 1 IMs underwent MRI scans at two, five, and 10 years after surgery. Those with gross-totally resected WHO grade 2 IMs were followed up annually for up to five years; if the five-year scan showed no recurrence, a final scan was scheduled at 10 years. Exceptions to this protocol were made on a case-by-case basis, especially when patients reported new neurological symptoms. IM recurrence was assessed by reviewing all postoperative radiology reports. If a patient was found to have an IM recurrence or a new unrelated intracranial tumor on follow-up MRIs, the follow-up was discontinued.

Intraoperative iatrogenic lesions (IILs) were assessed in patients with early postoperative in-hospital MRIs, which were performed only due to new postoperative neurological symptoms. IILs were defined as either acute ischemic lesions—new hyperintensities in diffusion-weighted imaging (DWI) sequences (Fig. 1E) [23] —or iatrogenic PTBE-like changes, which were considered as hyperintense FLAIR lesions around the resection cavity outside preoperative PTBE areas. To differentiate IILs from persisting PTBE, 3D Slicer was used to segment both. Specifically, preoperative PTBE (Fig. 1B), postoperative DWI hyperintensities (Fig. 1E), and IILs in in-hospital MRIs were segmented separately. Locations of preoperative and postoperative PTBE and IILs were assessed in follow-up MRIs to determine whether persisting changes represented PTBE or IIL.

Gadolinium-enhanced T1 images were used for preoperative intracranial meningioma (IM) segmentation. Fluid-attenuated inversion recovery sequence (FLAIR) was used to segment both pre- and postoperative peritumoral brain edema (PTBE). All hyperintensity surrounding the tumor or surgical site was considered as PTBE. (A) T1 + C segmented IM can be seen inside the white box. (B) Segmented preoperative PTBE can be seen in light gray inside the white box. (C) Example of small FLAIR hyperintensity around the surgical site. This small change was classified as persisting PTBE (in the white box). (D) 3D reformat of tumor segmentation in dark gray and surrounding PTBE in light gray. (E) Example of acute ischemia around the surgical cavity in early postoperative in-hospital diffusion weighted MR imaging (in the black box)

Given the study hospital’s follow-up protocol, we aimed to report the frequency of PTBE resolution for study patients at one and two years, as well as at last MRI follow-up after surgery. The percentage that PTBE resolved postoperatively was calculated by comparing the preoperative PTBE volume to the volume in the last follow-up MRI. The resolution percentage was calculated as: \(Resolution\>percentage\)\(= \>(1 - \over })*100\% \).

For patients with partial or complete PTBE resolution, we assessed whether EI was associated with PTBE resolution at last follow-up. Patients were grouped by preoperative EI into small PTBE (EI < 2), moderate PTBE (EI 2–3), and large PTBE (EI > 3).

Categorical variables were reported as counts and percentages. Normally distributed numerical data is presented as mean (standard error), while non-normally distributed data is shown as median (interquartile range). For PTBE association analyses, only patients with a reduction in PTBE volume were included, as increases are likely related to iatrogenic factors. The Welch t-test was used for two-category variables (sex, recurrence, WHO grade). For variables with more than two categories (edema location, EI, location, tumor laterality, histopathology), linear regression with the “lm” function was used to assess associations with PTBE resolution percentage, using the most frequent category as a reference. The model specification was: Y = β0 + β1 × Category1 + β2 × Category2 +…+ βn × Categoryn + ϵ. Patients with unspecified sex or histopathology were excluded from those specific analyses, but included in other analyses when appropriate. P-values below 0.05 were considered statistically significant. All analyses were conducted using RStudio version 2023.9.0.463 [24].