Abstract
Background/Aim: NF2-related schwannomatosis is a rare hereditary tumor predisposition syndrome, formerly known as Neurofibromatosis type 2 (NF2), and is characterized by the development of multiple schwannomas. The hallmark manifestation is the occurrence of bilateral vestibular schwannomas (VSs). Disease progression and clinical outcomes vary widely among patients, and conventional magnetic resonance imaging (MRI) metrics—such as tumor size—do not fully account for these differences. Radiomic analysis offers a quantitative approach to extract advanced imaging biomarkers that may capture tumor microstructure and growth behavior more accurately, potentially improving prognostication and individualized management in NF2.
Patients and Methods: A retrospective cohort of 32 patients with NF2 was analyzed, comprising 170 cranial MRI scans of 232 VSs (112 left, 120 right) acquired at different time points over the course of disease, with previously treated tumors excluded. Tumor growth was quantified as absolute and percentage growth rates per month. Radiomic features were extracted from segmented tumors and correlated with Common Terminology Criteria for Adverse Events (CTCAE) graded clinical findings. Hearing impairment was graded based on subjective function due to limited audiometric data.
Results: In this exploratory analysis, no Spearman |ρ| >0.40 correlations were observed between radiomics features and tumor growth rates, nor among clinical parameters (excluding depression and anxiety). Right-sided tumors were associated with significantly greater hearing impairment compared to left-sided tumors despite comparable volumes (p=0.030). Age-stratified analyses revealed distinct patterns: in younger patients (<30 years), fast-growing tumors displayed more homogeneous texture profiles, while in older patients (>30 years), rapid growth was linked to greater heterogeneity.
Conclusion: Radiomic profiling indicates that both tumor laterality and patient age influence the relationship between imaging features and clinical outcomes in NF2-associated VS. Homogeneity was linked to aggressive growth in younger patients, heterogeneity characterized progression in older patients. These findings suggest that radiomic biomarkers may complement volumetric measures and support individualized monitoring and treatment strategies in NF2.
Introduction
NF2-related schwannomatosis (NF2) is an autosomal dominant inherited disorder that predisposes affected individuals to the development of multiple neural tumors (1). The incidence is estimated at approximately 1 in 28,000, with a prevalence of about 1 in 50,000, numbers expected to increase due to advances in diagnostic techniques and an aging population (2, 3). The clinical characteristic feature of NF2 is the occurrence of bilateral vestibular schwannomas (VSs), which are present in nearly 90% of patients (4, 5).
Management of NF2 requires a careful balance between conservative surveillance and surgical intervention. In cases where vestibular schwannomas are not surgically accessible, systemic therapy with bevacizumab, an anti-VEGF monoclonal antibody, has also emerged as a treatment option to reduce tumor size and preserve hearing in selected patients (6). Radiotherapy may be considered when neither surgery nor medical therapy are feasible (7). All strategies, however, are associated with considerable risks (8-10). Although these risks are recognized, they have not yet been systematically investigated. The primary objective of therapy is to preserve neurological function and to maintain quality of life for as long as possible (11).
Age at first diagnosis represents an important prognostic marker in NF2. Patients with first diagnosis at a younger age typically show a more aggressive disease course, characterized by faster VS growth, greater tumor burden at baseline, and earlier deterioration of hearing function (Wishart phenotype) (12-14). Longitudinal NF2 cohorts have demonstrated that VS growth rates decline with advancing age (12, 13), while large-scale series and clinical overviews underscore the variability of tumor dynamics across different age strata (15, 16).
Radiomics enables the extraction of large sets of quantitative features from standard medical images, capturing tumor intensity, shape, and texture. These data can be correlated with clinical and molecular parameters to improve disease characterization, prognosis, and treatment planning, thereby supporting precision medicine (17-19).
Previous studies of sporadic VS have demonstrated that specific radiomic profiles can predict hearing decline (20) and may guide surveillance intervals. In our own prior work (21), we identified the radiomic feature wavelet-High–Low–High direction vector (HLH) - Gray Level Co-occurrence Matrix (GLCM)-Inverse Variance as a potential predictor of NF2-associated VS growth. Recent molecular studies have shown that pathogenic variants and allele loss in the NF2 and LZTR1 genes are prerequisite events for vestibular schwannoma development, yet even a higher mutational burden, although associated with increased structural heterogeneity on MRI, does not appear to determine tumor growth or clinical outcome (22).
These findings suggest that imaging-based tumor characteristics may capture biological variability beyond genetic alterations. Building on this concept, the present study applies radiomic analysis to quantify MRI-derived shape and texture features and to explore their associations with clinical parameters such as hearing loss and tumor size.
Patients and Methods
All analyses were performed on pseudonymized data, which were not traceable to individual patients. In accordance with the declaration from the Ethics Committee of the Hamburg Medical Council (Case Number: 2024-300549-WF), no concerns were raised regarding this research.
Study population. This retrospective analysis included 182 cranial magnetic resonance imaging (MRI) scans from 35 consecutive patients with NF2 who were treated at the Neurofibromatosis Outpatient Department of the University Medical Center Hamburg-Eppendorf between 2002 and 2020. Genetic testing revealed no mosaic variants in this cohort. The diagnosis of NF2 was established according to the criteria applicable at the time of the patients’ initial presentation (23). Two patients were excluded due to implausible growth trajectories; in two further cases, imaging of one side was omitted after neurosurgical intervention. MRI data from four patients were excluded following treatment with Bevacizumab or Everolimus.
Following exclusions, the analysis comprised 170 cranial MRIs with a total of 232 vestibular schwannomas (left: 112; right: 120) from 32 patients. The median age at the time of the first MRI was 31 years (range=2-67 years), and the cohort consisted of 20 females and 12 males.
Radiomics analysis. For details of the workflow for radiomics analysis please refer to our previous paper (21).
Tumor growth. Tumor growth dynamics were assessed using two measures: 1) Mean monthly absolute growth rate (AGRpM) – defined as the absolute change in tumor volume (mm3) per month, averaged across the individual follow-up interval; 2) Percentage Growth Rate per Month (PGRpM) – defined as the relative change in tumor volume per month, expressed as a percentage of the initial tumor volume at baseline MRI.
Clinical findings. In the Neurofibromatosis Outpatient Department of the University Medical Center Hamburg-Eppendorf, a pseudonymized registry is routinely maintained, documenting Common Terminology Criteria for Adverse Events (CTCAE)-graded severity of symptoms and clinical findings (24). This registry captures the typical functional impairments observed in patients with NF2.
The CTCAE was selected as the grading framework, as it represents the most widely accepted and standardized system for the assessment of symptom severity and functional impairments in oncology (25). For each CTCAE item, a five-point severity grading (Grade 1-5) is provided, with explicit definitions that reflect increasing clinical impact from mild manifestations to death. In total, 204 CTCAE reports from 29 patients were available, including 80 reports obtained within ±3 months of a corresponding radiomics-analyzed MRI. With respect to hearing impairment, only limited audiometric data were available, so that only subjective hearing function could be considered in the grading.
Statistical methods. Correlation analyses were conducted using Spearman’s rank correlation coefficient. The analyses between groups were performed with Welsh t-tests.
Results
Age and tumor growth. To analyze potential age-related influences on tumor characteristics, Spearman’s rank correlations were calculated. A significant negative correlation was found between patient age at the time of the first MRI and the AGRpM (ρ=−0.31; p=0.010). A significant negative correlation was also observed with the PGRpM at the time of the first MRI (ρ=−0.31, p=0.014). No significant correlation was found between age and initial tumor volume (shape_VoxelVolume) at the first MRI (ρ=0.17; p=0.194).
Correlations of clinical findings with shape based radiomics features. No correlation coefficients (|ρ| >0.40) were observed between tumor size or shape-based radiomic features (e.g., maximum tumor diameter, volume-to-surface ratio) and clinical findings.
Intercorrelations of clinical features. Spearman intercor-relation analysis of clinical findings revealed a strong positive correlation between depression and anxiety (ρ=0.62, p<0.001), as shown in Figure 1. No other correlations exceeded the threshold of |ρ| >0.30.
Heatmap illustrating Spearman’s rank correlation coefficients between CTCAE-graded clinical symptoms.
Laterality-dependent associations between tumor size and hearing loss. A Welch t-test of tumor volumes by side (left N=41, right N=37) showed no significant difference [t(−0.851), p=0.398]. For small- medium tumors (<8.0 cm3, 90.6% of all tumors), however, a significant difference between side and degree of hearing impairment was observed [t(−2.206), p=0.030, Figure 2].
Laterality-dependent association between tumor size and hearing loss. Comparison of hearing function between right- and left-sided vestibular schwannomas. For small to medium tumors (<8.0 cm3, 90.6% of all tumors), right-sided lesions were associated with significantly greater hearing loss (t=−2.206, p=0.030) despite comparable tumor volumes (t=−0.851, p=0.398).
Age, tumor homogeneity, and tumor growth. The radiomics feature Inverse Variance reflects local gray-level homogeneity, with lower values associated with greater intratumoral heterogeneity (26). In younger patients, tumors with fast growth exhibited significantly lower heterogeneity compared to slowly growing tumors [t(-3.265, p=0.003]. Conversely, in older patients, low heterogeneity was observed in the subgroup with slow tumor growth [t(2.953), p=0.006 Figure 3].
Age-dependent relationship between tumor texture homogeneity and growth rate. Radiomic feature GLCM Inverse Variance (reflecting local gray-level homogeneity) stratified by age group and tumor growth dynamics. In younger patients (<30 years), rapidly growing tumors exhibited greater homogeneity (lower Inverse Variance), whereas in older patients (>30 years), fast-growing tumors showed higher heterogeneity (higher Inverse Variance). Error bars indicate interquartile range (25 and 75%).
Discussion
Study limitations. This study has several noteworthy limitations. Being retrospective in nature, the generalizability of its findings is inherently constrained (27-29). Moreover, the modest sample size of 32 patients warrants cautious interpretation. Another important consideration is the 5-mm slice thickness of the segmented T2-weighted sequences, which may introduce partial volume effects and thus potential bias (30). To minimize this risk, each segmentation was cross-referenced with the preceding time point.
Age and tumor growth. In our NF2 cohort, patient age at first MRI was inversely associated with relative growth rates (AGRpM and PGRpM), while no relationship was observed with baseline tumor volume. This finding suggests that younger patients with NF2 may experience more aggressive tumor dynamics, consistent with earlier reports that younger age at onset is linked to higher growth activity and a more severe clinical phenotype (9, 12). The absence of a correlation between age and initial tumor volume emphasizes that baseline size at diagnosis is not a reliable marker of subsequent behavior in NF2. Rather, the effect of age appears to manifest in longitudinal kinetics rather than cross-sectional measurements, which aligns with prior observations that tumor burden and progression in NF2 are influenced more strongly by genetic background and age at onset than by presentation size (31).
Taken together, these findings highlight age at presentation as a potential prognostic factor in NF2-associated VS and support close surveillance of younger patients, even in cases with relatively small initial tumor volumes.
Correlations of clinical findings with shape based radiomics features. No correlation coefficients (|ρ| >0.40) were observed between tumor size or shape-based radiomic features and clinical findings. This suggests that macroscopic geometric properties, such as maximum tumor diameter or the volume-to-surface ratio, may have limited explanatory value for symptom burden in NF2-associated vestibular schwannomas. The absence of strong associations is consistent with prior work indicating that clinical manifestations are often determined less by absolute size than by tumor location, growth direction, and involvement of critical neurovascular structures (12, 31). Moreover, shape descriptors provide only coarse information on tumor geometry and do not capture microstructural heterogeneity or nerve compression patterns, which are likely to be more relevant for functional impairment. These findings highlight the importance of complementary radiomic texture features and longitudinal growth metrics, rather than static size- or shape-based measures alone, when attempting to link imaging characteristics with clinical outcomes.
Intercorrelations of clinical features. To our knowledge, no previous studies have systematically examined intercorrelations among clinical symptoms in NF2-associated VS. Existing literature primarily describes the frequency and sequence of symptom onset, with hearing loss typically representing the earliest manifestation and often accompanied by tinnitus or imbalance (7, 11). However, statistical relationships between symptom domains have not been reported. Our correlation analysis therefore provides novel insight into how distinct clinical features cluster in NF2. Among the intercorrelations of clinical findings, the strongest association was observed between depression and anxiety (ρ=0.62), reflecting the close relationship of these neuropsychiatric comorbidities. Moderate correlations were also observed between tinnitus and vestibular dysfunction, as well as between depression and gait disturbance. These findings highlight the multifactorial burden of NF2 and suggest that symptom domains may partially interact, rather than occurring in complete isolation.
Laterality-dependent associations between tumor size and hearing loss. As audiometric data were limited, grading of hearing impairment relied on subjective hearing function, based on patients’ self-reported hearing ability. The observed laterality-dependent difference in hearing impairment, with more pronounced dysfunction in right-sided tumors despite comparable volumes, suggests that factors beyond tumor size contribute to auditory decline. Functional neuroimaging and neurophysiological studies indicate hemispheric asymmetries in auditory processing, with the left auditory cortex more specialized for temporal resolution and speech-related features, and the right auditory cortex preferentially involved in spectral and pitch-related processing (32). Such lateralized cortical organization may render right-sided vestibular schwannomas more disruptive to specific auditory pathways, thereby amplifying functional impairment relative to left-sided lesions of similar size. These findings underscore the need for future studies integrating structural, functional, and radiomic analyses to clarify the mechanisms underlying laterality effects in NF2-associated VS.
Age, tumor homogeneity, and tumor growth. Age-stratified analysis showed that the association between tumor homogeneity and growth differed by age. In younger patients, fast-growing tumors were characterized by more homogeneous texture patterns (low values of GLCM Inverse Variance), whereas in older patients, rapid growth was linked to greater heterogeneity (high values of GLCM Inverse Variance).
The isolated significance of log-sigma-6-5-3D_glcm_ InverseVariance may be explained by its ability to emphasize regional intensity patterns at a biologically relevant scale. Log filtering highlights tumor microstructural heterogeneity by suppressing noise and enhancing structures within a defined frequency band, thereby stabilizing texture descriptors such as Inverse Variance. Other homogeneity features, operating at finer or broader scales or without filtering, may dilute these regional effects and thus fail to capture them robustly (33-35).
This age-dependent importance of radiomic homogeneity metrics should be interpreted cautiously. To date, there is no robust evidence that age modifies the association between radiomic homogeneity metrics (e.g., GLCM-based features) and phenotype in NF2-associated vestibular schwannoma. We are not aware of studies that explicitly model an age × homogeneity interaction in NF2 cohorts. Existing NF2 literature chiefly reports age- and genotype-related differences in growth dynamics and describes structural subtypes (e.g., cystic change) in long-standing or treated lesions, but it does not provide age-stratified analyses of MRI-derived homogeneity features. Accordingly, any age-dependent interpretation of radiomic homogeneity in NF2 should be considered hypothesis-generating and requires prospective validation.
Analogously though, in pediatric diffuse midline glioma, medium-scale MRI texture homogeneity at diagnosis predicted worse survival, underscoring that age-group and spatial scale can invert the prognostic meaning of ‘homogeneity’ (36), a concept consistent with our age-stratified findings in NF2.
The age-dependent association between tumor homogeneity and growth may reflect underlying biological differences. In younger patients, rapid growth likely arises from aggressive but relatively uniform tumor cell populations, consistent with prior evidence that younger age and truncating NF2 mutations are linked to faster progression and more severe phenotypes (37, 38). In contrast, fast-growing tumors in older patients may be driven by secondary changes such as cystic degeneration or fibrosis which have been described in long-standing or treated NF2-associated VS. These findings may suggest that distinct biological mechanisms underlie tumor progression in different age groups in NF2-associated VS. Taken together, these findings indicate that radiomic measures of homogeneity may provide an age-dependent biomarker of tumor aggressiveness and could help refine both monitoring strategies and treatment decisions in NF2-associated VS.
Conclusion
In this NF2-related schwannomatosis cohort, younger age at first MRI was linked to faster longitudinal tumor growth, while baseline tumor volume showed no age dependence. This suggests that age-related risk is primarily reflected in tumor growth kinetics rather than initial size, consistent with prior evidence that NF2 progression is more strongly influenced by genetic background and age at onset than by presentation volume.
No significant associations emerged between size- or shape-based radiomic features and clinical symptoms, indicating that macroscopic tumor geometry alone provides limited insight into functional outcomes. Despite comparable tumor volumes, right-sided vestibular schwannomas were associated with greater hearing impairment, supporting the influence of lateralized auditory processing and side-specific neural pathways over bulk tumor effects.
Age-stratified texture analysis revealed a potential effect modification: in younger patients, faster-growing tumors exhibited more homogeneous texture profiles, while in older patients, rapid growth corresponded to greater heterogeneity. The prominence of the log-sigma-6-5-3D_glcm_InverseVariance feature likely reflects its ability to capture biologically meaningful mid-scale intensity patterns.
Overall, these findings highlight the value of incorporating age- and texture-aware radiomic biomarkers into NF2 monitoring strategies. Radiomic profiling may complement conventional volumetric measures by providing noninvasive indicators of biological behavior and could support individualized surveillance and treatment planning in NF2-associated vestibular schwannomas. Future prospective studies are warranted to validate these hypothesis-generating observations.
Acknowledgements
The MRI series and anonymized clinical data were kindly provided by Prof. Victor-F. Mautner from the Department of Neurology, University Medical Center Hamburg-Eppendorf (UKE).
Footnotes
Authors’ Contributions
Study conception and design: Nina Boe, Johannes A. Koeppen, Victor-F. Mautner; Data collection: Victor-F. Mautner; Data analysis: Nina Boe, Johannes Koeppen; Results interpretation: Reinhard E. Friedrich, Victor-F. Mautner, Nina Boe, Johannes A. Koeppen; Drafting the manuscript: Nina Boe, Johannes Koeppen, Reinhard E. Friedrich, Said C. Farschtschi, Hanno S. Meyer; Manuscript revision and approval: All Authors.
Conflicts of Interest
SF has received speaker honoraria from Alexion and compensation for advice or lecturing from SpringWorks and Alexion not related to this study. All other Authors declare that they have no conflicts of interest.
Funding
This study was funded by the German lay organisation Bundesverband Neurofibromatose.
Artificial Intelligence (AI) Disclosure
All statistical analyses were performed without the use of artificial intelligence. ChatGPT (OpenAI) was employed solely to assist with literature searches and the translation of German text into English. Its output was used only in part and was carefully reviewed and verified by the Authors before inclusion.
- Received October 14, 2025.
- Revision received November 3, 2025.
- Accepted November 5, 2025.
- Copyright © 2026 The Author(s). Published by the International Institute of Anticancer Research.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.