Purpose: To provide early and localized GBM recurrence prediction, we introduce a novel post-surgery multi-parametric MR based support vector machine (SVM) method coupling with stem cell niches (SCN) proximity estimation.
Methods: This study utilized post-surgery MRI scans ~2 months before clinically diagnosed recurrence from 58 GBM patients. The main prediction pipeline consists of a proximity-based calculation step, to identify regions with high risks of recurrence, and an SVM classifier training step, to further leverage voxel-wise prediction in the high-risk regions (HRR). The HRRs were estimated using the weighted sum of inverse distances to two possible origins of recurrence - SCN (PS) and tumor cavity (PT). Subsequently, multi-parametric voxels (from T1, T1 GD, FLAIR, T2, ADC MR) within the HRR were grouped into the recurrent subregions (warped from the time of clinical diagnosis) and non-recurrent subregions, and feed into a radial basis function SVM classifier. 49 patients were used for training and validation, and 9 patients were used for testing. Besides SVMPE, an SVM classifier using tumor margin as input high-risk regions (SVMTM) was also trained for comparison.
Results: On pre-recur MRIs from 9 testing patients, the SCN proximity based HRRs achieved a maximized inclusion of recurrence, with a recall of 0.97. The fine-tuned voxel-wise prediction from the SVMPE classifier reached a recall of 0.84, a precision of 0.74, and an F1-score of 0.77. In contrast, performance from SVMTM was consistently lower (recall 0.66, precision 0.59, and F1 0.60).
Conclusion: We introduced a novel voxel-wise early prediction method for brain GBM recurrence based on clinical follow-up MR scans using an SCN proximity estimation coupled SVM framework-SVMPE. The result shows promise in localizing subclinical traces of recurrence 2-month ahead of clinical diagnosis and may be used to guide early retreatment to a smaller target volume and individualized prescription.