To compare 1D (linear) tumor volume calculations and classification systems with 3D-segmented volumetric analysis (SVA), focusing specifically on their effectiveness in the evaluation and management of NF2-associated vestibular schwannomas (VS). VS were clinically followed every 6 months with cranial, thin-sliced (< 3 mm) MRI. We retrospectively reviewed and used T1-weighted post-contrast enhanced (gadolinium) images for both SVA and linear measurements. 3D-SVA was performed manually or combined with semiautomated segmentation by using axial planes. The maximum linear dimensions (MLD) were determined in three dimensions (anteroposterior, transverse, and craniocaudal planes) using axial and coronal planes. The MLD was cubed (MLD3), and orthogonal analysis (OA) was derived to establish comparability with the SVA. The Hannover and Koos classification was used to depict the size ratio in each MRI and tumor. A linear regression model was performed to compare 1D/classification systems to SVA, and the percentage deviation change of MLD3 and OA to SVA was established using a one-way multivariate variance analysis. 2586 SVA and 10344 linear measurements were performed in a cohort of 149 NF2 patients and 292 associated VS. All measurement techniques (MLD3, OA, KOOS, and Hannover) significantly (and strongly, r2 > 0.5) correlated with SVA (p < 0.001). The OA showed an even stronger positive correlation than the MLD3 to SVA. Smaller classified tumors (T1/T2, K1/K2) exhibited a low-moderate positive correlation (r2 = 0.23–0.44) compared to medium-sized (T3, K2/3) and large tumors (T4, K4; r2 = 0.54–0.76). Pre- and postoperative MLD3 and OA statistically significantly predict SVA (p < 0.001), but the postoperative correlation was weaker, particularly for MLD3 to SVA values. All analyses showed a large scatter range. In the percentage deviation analysis of MLD3 and OA from SVA, small tumors (K1/K2, T1/T2) were overestimated. Compared to the SVA, the MLD3 and especially the OA are a time-saving alternative for monitoring the tumor volume of NF2-associated VS. However, the scatter range in small/surgically reduced tumors is enormous. For this reason, they are not recommended for monitoring off-label therapy with Bevacizumab or for treatment decisions depending on a precise assessment of tumor volume and growth. Developing deep learning-based volume determinations in the future is essential to reduce SVA’s time intensity.
Read full abstract