[18F]BBPA PET/CT and MRI were performed within 1 week on separate days. For [18F]BBPA PET/CT, a dose of 3.7 MBq (0.1 mCi)/kg [18F]BBPA was intravenously given, and a PET/CT scan was acquired using a Biograph mCT Flow 64 scanner (Siemens, Germany) 30 min after injection. The PET image was transferred into an SUV map that was normalized by body weight and decay factor. For MRI, contrast-enhanced T1-weighted MRI (matrix 256 × 256, slice thickness 1 mm, gadolinium chelate 0.1 mmol/kg) and T2-weighted MRI (matrix 256 × 256, slice thickness 5–6 mm) were acquired from a 3.0 T Discovery MR750 scanner (GE, USA).

[18F]BBPA PET/CT image and T2-weighted MRI were co-registered to the thin-slice contrast-enhanced T1-weighted MRI to unify the origin and direction of images, allowing the same region of interest (ROI) refers to identical area in different image modality.

Patients that were suspected to have primary or metastatic brain tumors were enrolled under the following criteria: (1) age ≥ 18 years; (2) Karnofsky Performance Score (KPS) ≥ 80; (3) suspected to have malignant gliomas or metastatic brain tumors based on medical history, clinical and radiological evaluation; (4) no contradictions for PET/CT and MRI scan. The pathological diagnosis was established by two neuropathologists according to the 2021 WHO classification for central nervous system tumors [24]. The therapeutic strategies, including but not limited to, surgery, radiotherapy, pharmacological treatment, or close imaging follow-up, were determined by a multi-disciplinary team after PET/CT and MRI scans.

Three spherical reference regions of interest (ROIref) with a diameter of 1 cm were manually placed on the contralateral area (mirroring the position of the tumor) to calculate the maximum and mean SUV of the normal brain (generating Nmax and Nmean, respectively) [22].

The ROI of the lesion was delineated by the definition of gross total resection (GTR) for brain tumors, which includes the contrast-enhanced region for significantly contrast-enhanced tumors or the region with abnormal T2-weighted signal for non-significantly contrast-enhanced tumors. The ROI was semi-automatically delineated and manually revised by a neurosurgeon on the thin-slice T1-weighted MRI using 3D Slicer (4.11.2, www.slicer.org). The ROI was subsequently applied to the co-registered BBPA PET images for feature calculation and histogram analysis.

Five traditional quantitative parameters, namely SUVmax, SUVmean, metabolic tumor volume (MTV), total lesion activity (TLA) and tumor-to-normal brain ratio (T/N ratio), were calculated [25]. SUVmax and SUVmean represent the maximum and mean SUV of ROI, while MTV and TLA calculate the volume and total radioactivity inside ROI. The T/N ratio was calculated as the ratio of SUVmax and Nmax.

The SUV of each voxel within ROI was documented as a number series, and a histogram was plotted to visualize the voxel value distribution. Skewness and tendency were defined to reflect the histogram characteristics:

$$}=\frac}}_}}}\sum_}=1}^}}_}}}(}(})-\overline}}^}}}_}}}\sum_}=1}^}}_}}}(}(})-\overline}}^}\right)}^}$$

where X refers to all voxel values included in the ROI, \(}}_}}\) refers to the number of voxel within ROI.

where SUVmean and SUVmedian refer to the mean and median SUV value within ROI.

Images were processed and segmented on 3D slicer (4.11.2, www.slicer.org). The Wilcoxon rank-sum test was applied to evaluate whether a parameter was significantly different in distinct circumstances. Statistical analysis were performed using Python (3.8.5, www.python.org) and R (4.0.4, www.r-project.org).