Early Detection of Glioblastoma by Quantitative MRI Radiomics in a Xenograft Model and Patients

Introduction: Glioblastoma (GBM) is the most fatal primary brain tumor in adults. The mainstay imaging technique for brain cancer diagnosis, assessing treatment response, and determining progression is magnetic resonance imaging (MRI). Unfortunately, conventional MRI can’t detect GBM beyond a certain tumor mass. Here we describe a quantitative MRI radiomics (QMR) signature that detects key features within conventional MRI resulting in early tumor cells detection ahead of conventional MRI.

Methods: Orthotopic GBM mouse models were generated by injecting patient-derived glioma stem cells (GSCs) intracranially into SCID-Ɣ-IL2 mice. GSCs lines were tagged with GFP or RFP along with a luciferase reporter gene. Bioluminescent imaging (BLI) was used to detect light signals. Serial MRI scans were obtained post-injection. QMR was applied to the mice MRI scans and a QMR signature was developed with the aim of early detection of GBM cells. We then applied this QMR signature to human MRI scans to test for early detection of GBM cells in humans.

Results: Mice bearing luciferase tagged tumors initially demonstrated a decreased BLI signal (engrafting phase; day 8 post-injection, MRI negative), followed by stabilization of luciferase signal and linear increase (linear phase; day 9-21, MRI negative) and subsequently, exponential increase in luciferase signal (exponential phase; day 22-43, MRI positive) with only now clear tumor being present on conventional MRI. QMR was used to create a radiomic biomarker signature for early prediction of MRI negative tumor during the linear phase. We then applied the biomarker signature to Glioblastoma patient MRI scans and detected patterns of future loco-regional tumor recurrence (P < 0.005) ahead of conventional MRI demonstrating a clear signal abnormality.

Conclusion : Our proposed QMR signature successfully predicted Glioblastoma stem cell presence ahead of conventional MRI in tumor-bearing mice and patients. Quantitative radiomics may augment conventional MRI towards the detection of early cancer cell presence in the brain.