Voxelwise Prediction of Recurrent High-Grade Glioma via Proximity Estimation-Coupled Multidimensional Support Vector...

Area of Science:

• Neuro-oncology
• Medical imaging analysis
• Machine learning in medicine

Background:

• Glioblastoma (GBM) recurrence poses a significant challenge for patient prognosis.
• Early and localized prediction of GBM recurrence is crucial for effective treatment planning.
• Current methods often lack the precision for timely detection of subclinical recurrence.

Purpose of the Study:

• To develop and validate a novel postsurgery multiparametric magnetic resonance imaging (MRI)-based support vector machine (SVM) method coupled with stem cell niche (SCN) proximity estimation for early and localized glioblastoma (GBM) recurrence prediction.
• To identify high-risk regions (HRRs) indicative of impending GBM recurrence.
• To enable voxelwise prediction of recurrence approximately 2 months before clinical diagnosis.

Main Methods:

• Utilized postsurgery MRI scans from 50 recurrent GBM patients, acquired ~2 months before clinical recurrence diagnosis.
• Developed a prediction pipeline combining a proximity-based estimator (identifying SCN and tumor cavity proximity) and an SVM classifier (SVMPE).
• Employed multiparametric MRI data (T1, T1-contrast-enhanced, FLAIR, T2, ADC) within HRRs for voxelwise prediction, with random 40% training and 60% testing split.

Main Results:

• The SVMPE classifier achieved a recall of 0.80, precision of 0.69, and F1-score of 0.73 on 2-month prerecurrence MRI scans from 30 test patients.
• Demonstrated a mean boundary distance of 7.49 mm in predicting recurrence.
• Exploratory analysis revealed spatially consistent but smaller subclinical clusters and increased T1-contrast-enhanced and apparent diffusion coefficient values over time.

Conclusions:

• A novel voxelwise early prediction method, SVMPE, for GBM recurrence based on clinical follow-up MR scans was demonstrated.
• The SVMPE method shows promise in localizing subclinical GBM recurrence traces up to 2 months prior to clinical diagnosis.
• This approach may facilitate personalized early salvage therapy strategies for GBM patients.