MRI Radiomic Features Predict Peritumoral Brain Edema Resolution Post-Meningioma Resection
Overview
This study demonstrates that radiomic analysis of preoperative MRI can differentiate reversible peritumoral brain edema (PTBE) from irreversible gliosis in intracranial meningioma patients. Using machine learning models on radiomic features extracted from T2/FLAIR sequences, the authors developed a preliminary classifier predicting PTBE resolution at one year after gross total tumor resection.
Background
Intracranial meningiomas are the most common primary brain tumors, often presenting with peritumoral brain edema (PTBE) visible on T2/FLAIR MRI sequences. PTBE is associated with worse clinical outcomes including seizures, hemorrhage, and longer hospital stays. However, some T2/FLAIR hyperintensities may represent gliosis rather than edema, which does not resolve postoperatively. Differentiating edema from gliosis preoperatively remains challenging but is critical for predicting neurological recovery after surgery. Radiomics and machine learning offer advanced quantitative imaging analysis to address this diagnostic gap.
Data Highlights
Parameter
Value/Description
PTBE incidence in IMs
38–67%
PTBE volume resolution threshold
80% reduction at 1 year post-op
Gliosis persistence post-op
97% of gross totally resected IMs
Substantial PTBE resolution observed
~33% of patients
Imaging sequences used
T1CE, T2, FLAIR
Radiomic feature extraction tool
PyRadiomics with IBSI compliance
Machine learning environment
Python 3.10, Azure ML, xgboost
Key Findings
PTBE is present in up to two-thirds of meningioma cases and correlates with adverse clinical outcomes.
Preoperative T2/FLAIR hyperintensity includes both reversible edema and irreversible gliosis, which are indistinguishable by visual MRI assessment alone.
Radiomic features extracted from tumor and peritumoral regions on preoperative MRI can be used to predict PTBE resolution after surgery.
Machine learning models trained on these radiomic features achieved preliminary classification of edema resolution status at one year post-resection.
Approximately one-third of patients showed substantial (>80%) PTBE volume resolution, while gliosis-like changes persisted in most cases.
Clinical Implications
Radiomic analysis combined with machine learning offers a promising noninvasive tool to preoperatively distinguish reversible edema from permanent gliosis in meningioma patients. This differentiation can improve surgical planning and patient counseling by better predicting postoperative neurological recovery and edema resolution. Incorporating radiomics into routine MRI assessment may enhance personalized management strategies for intracranial meningiomas.
Conclusion
Radiomic features derived from preoperative MRI provide valuable predictive information regarding PTBE resolution following meningioma resection. This approach addresses a critical unmet need in differentiating edema from gliosis, potentially improving clinical outcomes through tailored treatment planning.
References
Suplementary Table 1, CLEAR guidelines, 2020 -- Study methodology adherence
Previous publication by authors, 2022 -- Imaging protocols and study population details
