# MRI-Based Prediction of Vestibular Schwannoma: Systematic Review

## TL;DR

This study reviews MRI-based biomarkers for predicting the growth and treatment response of vestibular schwannomas, finding potential but noting limitations in current evidence.

## Contribution

A systematic review of MRI-based biomarkers for predicting vestibular schwannoma growth and treatment response, highlighting their exploratory potential and current limitations.

## Key findings

- Texture analysis metrics like kurtosis and GLCM features showed AUCs of 0.65–0.99 for predicting tumor growth.
- Signal intensity ratios on gadolinium-enhanced T1-weighted images achieved 100% sensitivity and 93.75% specificity for growth prediction.
- Perfusion MRI parameters differentiated growing from stable tumors with AUCs up to 0.85.

## Abstract

A vestibular schwannoma is a benign tumor that can behave very differently across patients. While some tumors remain stable for many years, others may grow and require treatment. At present, clinicians rely on repeated MRI scans to monitor these tumors, but it is still difficult to determine early on which tumors are likely to grow. The objective of our research was to review and summarize existing studies on MRI-based biomarkers that might improve this prediction. These biomarkers include texture features, signal intensity ratios, perfusion parameters, and apparent diffusion coefficients. We found that these imaging features show promise for identifying tumors that are more likely to grow or to respond well to therapy. However, the available evidence is heterogeneous, largely based on small single-center cohorts, and lacks external validation. Therefore, these MRI-based markers should currently be regarded as exploratory tools rather than predictors ready for routine clinical use.

Background: The vestibular schwannoma (VS) is the most common cerebellopontine angle tumor in adults, exhibiting a highly variable natural history, from stability to rapid growth. Accurate, the non-invasive prediction of tumor behavior is essential to guide personalized management and avoid overtreatment or delayed intervention. Objective: To systematically review and synthesize the evidence on MRI-based biomarkers for predicting VS growth and treatment responses. Methods: We conducted a PRISMA-compliant search of PubMed, EMBASE, and Cochrane databases for studies published between 1 January 2000 and 1 January 2025, addressing MRI predictors of VS growth. Cohort studies evaluating texture features, signal intensity ratios, perfusion parameters, and apparent diffusion coefficient (ADC) metrics were included. Study quality was assessed using the NOS (Newcastle–Ottawa Scale) score, GRADE (Grading of Recommendations, Assessment, Development and Evaluation), and ROBIS (Risk of Bias in Systematic reviews) tool. Data on diagnostic performance, including the area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, specificity, and p value, were extracted and descriptively analyzed. Results: Ten cohort studies (five retrospective, five prospective, total n = 525 patients) met the inclusion criteria. Texture analysis metrics, such as kurtosis and gray-level co-occurrence matrix (GLCM) features, yielded AUCs of 0.65–0.99 for predicting volumetric or linear growth thresholds. Signal intensity ratios on gadolinium-enhanced T1-weighted images for tumor/temporalis muscle achieved a 100% sensitivity and 93.75% specificity. Perfusion MRI parameters (Ktrans, ve, ASL, and DSC derived blood-flow metrics) differentiated growing from stable tumors with AUCs up to 0.85. ADC changes post-gamma knife surgery predicted a favorable response, though the baseline ADC had limited value for natural growth prediction. The heterogeneity in growth definitions, MRI protocols, and retrospective designs remains a key limitation. Conclusions: MRI-based biomarkers may provide exploratory signals associated with VS growth and treatment responses. However, substantial heterogeneity in growth definitions and MRI protocols, small single-center cohorts, and the absence of external validation currently limit clinical implementation.