Validation of SUV thresholds in [¹⁸F]SiTATE PET/CT for accurate meningioma segmentation

Abstract

Purpose: Somatostatin receptor (SSTR)-targeted PET/CT provides valuable clinical insights beyond standard imaging in meningioma patients. Due to its excellent diagnostic capabilities and favorable logistics, the ¹⁸F-labeled SSTR-targeting peptide SiTATE is increasingly in demand. We aimed to validate a recently proposed standard uptake value (SUV) threshold for accurate meningioma delineation in a clinically diverse patient cohort, including complex anatomical locations and lesions with prior surgical intervention.

Methods: Consecutive patients with known or suspected meningioma who underwent [¹⁸F]SiTATE PET/CT and contrast enhanced cerebral MRI were included. Lesions were semi-automatically segmented on PET images using an individualized minimal SUV (SUV_min) within a manually defined volume of interest. Correlative CT and MRI images were used to refine segmentations for each lesion, identifying the optimal lesion-specific SUV_min to accurately capture the true volume of the meningioma. All lesions were additionally segmented using the recently proposed threshold of 4.0, and resulting volumes were compared.

Results: 61 patients with 109 lesions were analyzed: 40 (37%) extraosseous, 32 (29%) partial trans-osseous, and 37 (34%) predominantly intraosseous. The median optimal SUV_min for lesion delineation was 4.2. Osseous involvement did not significantly affect the median SUV_min (p = 0.1). Individualized SUV volumes showed excellent absolute agreement with those obtained using the fixed threshold of 4.0 (ICC[A,1] = 0.967; 95% CI: 0.952-0.977; p < 0.0001). However, 17 lesions (SUV_max < 4.2) were not captured by the fixed threshold.

Conclusion: The proposed SUV threshold of 4.0 showed promising results, supporting its suitability for clinical practice. Although limitations were evident, with 16% of lesions - primarily very small - showing reduced uptake and therefore not captured by this threshold, the study underscores its applicability in clinical practice.

Keywords: Meningioma; PET/CT; Segmentation; SiTATE; Threshold.

Fig. 1

Example of SUV-thresholding for an extraosseous meningioma located in the right temporal region, prior to any treatment. The images are displayed in four columns, showing different modalities at the same axial level. From left to right: CE-CT, PET scan, fused PET/CT image, and corresponding CE-MRI on the far right

Fig. 2

PET images showing 3D segmentations of the same extraosseous meningioma presented in Fig. 1, using an individualized threshold of 4.5 (red contour) and a fixed threshold of 4.0 (blue contour). The substantial overlap between segmentations demonstrates a high degree of concordance between the two methods, with only minimal differences in segmented volumes (39.3 ml vs. 37.2 ml)

Fig. 3

Distribution of Standardized Uptake Values (SUV) across different anatomical locations of meningiomas, including the median SUV_max and SUV_mean for both total and intraosseous parts of meningiomas. Osseous segmentation was performed using HU-thresholding (300–3000 HU) within the segmented volume defined by the individualized SUV_min. Each boxplot illustrates the variability and central tendency of SUV values at distinct anatomical sites, highlighting differences in uptake based on the meningioma location. The notches in the boxplots represent the confidence interval around the median, providing a visual indication of the variability in the data. A non-overlapping notch between two boxplots suggests a statistically significant difference in the medians of the two groups

Fig. 4

(A) Median optimal SUV threshold values (SUV_min) for the delineation of meningioma lesions with different extensions, alongside median SUV_mean values for non-meningioma structures in the subarachnoid space and the unaffected bone of the contralateral side. This comparison did not reveal a significant difference in SUV_min between lesions with and without osseous involvement (Mann–Whitney U test, p = 0.1). (B) Overall SUV metrics

Fig. 5

Scatter plot comparing lesion volumes for each individual lesion segmented using the individualized optimal SUV_min based on morphological comparison (y-axis) with volumes obtained using a fixed SUV cut-off of 4 (x-axis). Each blue dot represents an individual observation. The dashed red line represents equality, indicating identical values for both volume measurements. The intraclass correlation coefficient (ICC[A,1]) was 0.967 (95% CI: 0.952–0.977), indicating excellent agreement between the two methods. However, for very small lesions with volumes < 1 cm³, a divergence between the volumes becomes apparent