Standard Deviations of MR Signal Intensities Show a Consistent Trend during Imaging Follow-Ups for Glioblastoma Patients when Corrected for Non-Biological Heterogeneity Due to Hardware and Software Variation

Authors

Subhendra N. Sarkar, CUNY New York City College of TechnologyFollow
Rafael Rojas, Harvard University
Evans Lespinasse, CUNY New York City College of TechnologyFollow
XiangFu Zhang, CUNY New York City College of TechnologyFollow
Ruth Zeron, CUNY New York City College of TechnologyFollow

Document Type

Article

Publication Date

1-2023

Abstract

Introduction

Glioblastoma multiforme (GBM) has a poor prognosis in spite of advanced MRI guided treatments today. Routine MRI using conventional T₁ or advanced permeability based MRI of GBM often does not adequately represent changing tumor phases or overall survival. In this work, region of interest (ROI) based tissue MR standard deviation (SD) is demonstrated as an important MRI variable that could be a potential biomarker of GBM heterogeneity and radioresistance.

Materials and methods

MRI characterization is often qualitative and lacks reproducibility. Using standardized MRI phantoms we have normalized retrospective records of 12 radioresistant GBM patients that underwent radiation therapy (RT) with concomitant and adjuvant temozolomide (TMZ) chemotherapy followed by serial MR imaging with gadolinium contrast.

Results and discussion

We have identified key variables like hardware, software and protocol variation and have standardized those using test phantoms at five MR systems. We suggest GBM growth during the treatment period can be linked to normalized MRI signal and its fluctuations from session to session and from magnet to magnet by using an ROI derived standard deviation that corresponds to heterogeneity of the tumor MRI signal and changes in magnetic susceptibility. The time period observed in our patient group for peak standard deviations is approximately halfway through the tumor course and may correspond to a growth of more aggressive MES subtype of cells. To model the GBM heterogeneity we performed in vitro T₁ weighted inversion recovery MRI experiments at 3 T for porous media of silicate particles in 1% aq solution of Gadavist and linked SD with particle size and local gadolinium volume within porous media. Such in vitro models mimic the increased SD in radioresistant GBM and as a novel contribution suggest that finer texture with high surface area might arise approximately halfway through the overall survival duration in GBM.

Conclusion

Standard deviation as a measure of magnetic susceptibility may be collectively linked to the changes in texture, cell fractions (biological) and trapped contrast media (vascular as well as artifactual consequences) and should be evaluated as a potential biomarker of GBM aggressiveness than the overall MRI signal intensity from a GBM.

Comments

This work was originally published in Clinical Neurology and Neurosurgery, available at https://doi.org/10.1016/j.clineuro.2022.107553

This work was made possible with undergraduate student research stipends from City University of New York to Xiang Fu Zhang and Ruth Zeron under Emerging Scholars program.