This retrospective study was approved by the Ethics Review Committee of our institution (2023 A-169) and the need for informed consent was waived. Preoperative clinical, MRI, and postoperative pathological datas of patients with histopathologically confirmed diagnosis of meningioma from January 2019 to June 2022 in our hospital were retrospectively analysed. Inclusion criteria: (1) resection biopsy specimens were histopathologically confirmed to be meningiomas and immunohistochemical staining for CD8 was available; (2) all patients underwent meningioma resection 1 week after MRI; (3) complete clinical, imaging, and pathological data. Exclusion criteria: (1) patients received radiotherapy, targeted therapy, or other treatments before preoperative MRI scanning; (2) incomplete image sequences or failure to meet the diagnostic requirements; (3) patients with tumor diameter < 1 cm; (4) patients lost to follow-up; and (5) unable to obtain tissue specimens after surgical resection. Finally, a total of 102 patients with meningiomas were included in this study. The flow diagram of patient inclusion and exclusion as showed in Supplementary Material Fig S1.

The primary endpoint was tumor recurrence, defined as the time interval between initial diagnosis (surgery date) and recurrence date. All enrolled patients were followed-up with MRI scans until August 31, 2023. Recurrence was defined as a new lesion in the operated area or a > 10% increase in residual lesion volume on MRI T1C at any time during the follow-up period; otherwise, the patient was categorized as non-recurrence.

All enrolled patients underwent preoperative MRI examinations (T1-weighted image (T1WI), T2-weighted images (T2WI), DWI, and T1C), using a 3.0-T MRI Scanner (Siemens Verio, Erlangen, Germany) with an 8-channel head coil. The scanning parameters were as follows: (1) gradient echo sequences: T1WI ( repetition time [TR] = 550 ms, time to echo [TE] = 11 ms), slice thickness 5 mm, slice spacing 1.5 mm, field of view (FOV) 260 mm × 260 mm, and matrix 269 × 384; (2) turbo spin echo: T2WI (TR = 2200 ms, TE = 96 ms), echo time 10 ms, and two excitations, and matrix 256 × 256; (3) DWI: (TR = 4500 ms, TE = 102 ms), slice thickness 5 mm, slice spacing 1.5 mm, FOV 260 mm × 260 mm, and matrix 160 × 160, and b values of 0 and 1000 s/mm2; and (4) T1C: gadolinium diethyltriaminopentylacetate (0.1 mmol/kg) was injected through the elbow vein at a rate of 3 mL/s with the same scanning parameters as those for T1WI, and axial, coronal, and sagittal images were obtained.

The conventional MRI features of all enrolled meningioma patients were independently evaluated by two radiologists (radiologists 1 and 2, with 10 and 8 years of experience in diagnostic neuroradiology, respectively) using a double-blind method to assess the following conventional MRI features: (1) tumor location: skull base, non-skull base; (2) necrosis: present, absent; (3) tumor enhancement: homogeneous, inhomogeneous; (4) tumor shape: round, lobulation; (5) tumor maximum diameter; (6) tumor volume (VT = 4πabc/3); (7) peritumour edema diameter; (8) peritumour edema volume (VE = VT2-high - VT); and (9) edema index (EI = (VE + VT)  / VT). Two experienced neurosurgeons, unaware of the imaging datas, retrospectively analysed the Simpson grading of the patients who underwent meningioma resection from the PACS operative records.

Preoperative ADC images of all enrolled patients were obtained from DWI images after post-processing, and all ADC images were imported into MaZda software (version 4.6, Technical University of Lodz, Institute of Electronics, Łódz, Poland, http://www.eletel.p.lodz.pl/mazda/) after adjusting the window width and window position. Then the largest level of the tumor was selected on the axial ADC images, and the tumor boundary was determined by T1C images, and ROIs were manually sketched along the edges of the tumor on the axial ADC images by the two radiologists mentioned above. In order to better assess the heterogeneity of the tumor, ROIs included cystic and hemorrhage areas of the tumor, but excluded peritumoural edema areas. The sketched ROIs were filled in red and the following parameters were automatically generated: mean, variance, skewness, kurtosis, perc.01%, perc.10%, perc.50%, perc.90%, and perc.99%. Disagreements were resolved by negotiation when they occurred, and the final result was the average of two radiologist measurements.

CD8 immunohistochemical staining were performed in 102 patients with meningioma after surgical resection. Paraffin specimens of postoperative meningiomas were cut into 4 μm sections. The CD8 antibody (RMA-0514, 1:100; Maixin) was used for immunohistochemical staining, revealing brownish-yellow cell membranes that suggested the positive surface expression of CD8 protein. All CD8-stained sections were scanned as whole-slide images (WSI) using the Motic EasyScanner digital section scanning system and viewed in DSAssistant software. For each CD8 WSI image, three different high-power fields (HPFs, 400× magnification) were randomly selected as ROIs. Segmented ROI images were analyzed by a neuropathologist in Image J software (version ImageJ2, USA). Final values for CD8 + T cells were the average of three ROIs and expressed as percentages (%), as shown in Figs. 1 and 2.

The WSI was checked by DSAssistant software, and 3 hot spots ROI images of CD8 + T cells IHC maps were randomly divided

The percentage of CD8 + T cells was obtained by CD8 immunohistochemical staining and gray level transformation and threshold adjustment

Immunohistochemical analysis of Ki-67 was performed using a monoclonal mouse anti-human Ki-67 antibody; brownish-yellow nuclei were considered Ki-67 positive. Regions with the highest concentration of Ki-67-positive cells (number of Ki-67 antibody positives per 1,000 tumor cells) were selected for evaluation. The Ki-67 proliferation index (PI) was calculated as the number of positive cells/total cell count.

MedCalc 19.1 (MedCalc, Mariakerke, Belgium) and R software (version 4.2.1; http://www.Rproject.org were used for statistical analysis. Categorical variables were expressed as number of cases (percentage) and were tested using the chi-square test or Fisher’s exact test. The Shapiro-Wilk test was used to test the normal distribution of continuous variables. Conventional MRI features and histogram parameters conforming to the normal distribution were expressed as mean ± standard deviation, and the differences between the groups were compared by the independent samples t-test. The non-normally distributed were expressed using the median ± quartile spacing, and the differences between the groups were compared by the Mann-Whitney U test. A multivariate COX proportional risk model was used to identify independent risk factors for meningioma recurrence and a nomogram was constructed, and validated the model to predict recurrence of WHO grade 1 and WHO grade 2 meningiomas. Recurrence-free survival curves were plotted using the Kaplan-Meier method, and differences between curves were analysed by log-rank tests. The probability of recurrence at 1-year and 2-years was assessed by plotting the receiver operating characteristic (ROC) curve. Interobserver agreement of conventional MRI characteristics and ADC histogram parameters was assessed using the intraclass correlation coefficient (ICC), with good interobserver agreement at ICC > 0.75. p < 0.05 was considered a statistically significant difference.