IMG-95. TTFields compliance reflects lower cellular tumor burden within contrast enhancement and FLAIR hyperintensity in glioblastoma patients

Abstract Tumor treating fields (TTFields) are a treatment available for recurrent glioblastoma (GBM), utilizing four transducer arrays placed on the scalp that produce alternating electric fields to disrupt cell division. We have a published machine learning algorithm that uses MRI data and autopsy tissue to generate radio-pathomic maps (RPMs) of cell density and tumor probability. In this study, we tested the hypothesis that GBM patients treated with TTFields therapy would have lower cell density and tumor probability within contrast enhancement (CE) and FLAIR hyperintensity (FH) compared to TTFields-naïve patients. 59 GBM patients from the MCW Neuro-oncology Brain Bank were split into three groups: eligible TTFields users (≥75% compliance for ≥90 days, n=17), low-usage TTFields users (<75% compliance or <90 days, n=21), and controls (no TTFields treatment, n=21). For each patient’s final MRI, RPMs were generated using T1, T1C, FLAIR, and ADC as input, and masks of CE and FH were annotated on the T1C and FLAIR. Two-tailed t-tests were used to compare average cell density and tumor probability within these masks between groups. Eligible TTFields patients showed significantly lower cell density (p=0.012) and tumor probability (p=0.002) within FH as well as significantly lower cell density within CE (p=0.012) compared to controls in the MRI scan closest to death. Low-usage TTFields patients showed significantly lower tumor probability within FH (p=0.027) compared to controls. These findings demonstrate that compliant TTFields treatment is associated with reduced cell density and tumor probability within both enhancing and non-enhancing tumor regions. Partial compliance or short-term use may harbor some treatment benefit, as even low-usage patients had reduced tumor probability within FH. However, full compliance resulted in patients demonstrating this effect, along with significantly reduced cell density in both CE and FH. Additionally, this result highlights the potential for treatment response assessment using radio-pathomic mapping in non-enhancing tumor margins.