2699 - Examining the Effects of Whole Brain Radiation Therapy and Stereotactic Radiosurgery on Brain Aging by Surface-Based Deep Learning
Presenter(s)
S. Zhao1, J. Yue2, Z. Fan3, Q. S. Zhang4, J. Rajamanickam5, V. Gondi6, C. G. Robinson1, T. L. S. Benzinger5, H. Li1,3, and N. Rammohan1; 1Department of Radiation Oncology, WashU Medicine, St. Louis, MO, 2Department of Electrical and Computer Engineering, Vanderbilt University, Nashville, TN, 3Department of Bioengineering, University of Illinois Urbana-Champaign, Champaign, IL, 4Department of Radiation Oncology, McGaw Medical Center of Northwestern University, Chicago, IL, 5Department of Radiology, WashU Medicine, St. Louis, MO, 6Northwestern Medicine Cancer Center Warrenville and Proton Center, Warrenville, IL
Purpose/Objective(s): Currently, there is no effective imaging biomarker for radiation-induced neurotoxicity. Our group previously established a graph convolutional neural network (CNN) model utilizing surface morphology from brain MRIs which demonstrated global structural aging following whole brain radiation therapy (WBRT). In this retrospective study, we examine the effect of treatment volume on structural brain changes by assessing brain aging following WBRT and stereotactic radiosurgery (SRS) compared to healthy controls. We hypothesize that both WBRT and SRS accelerate brain aging compared to healthy controls.
Materials/Methods: Pre-RT and follow-up MRIs for patients who underwent WBRT and SRS without radiological progression of disease were included. Brain MRIs from healthy controls were acquired from publicly available repositories. Open-source software surface anatomy data were input into the CNN model. Aging pace of the whole brain, cortex, hippocampus and combined subcortical structures was estimated with a linear mixed-effects model: ?Age = f(?Scan) + ?, where ?Age is the difference in model-predicted age and actual age between baseline and follow-up, f is the estimated aging pace for WBRT or SRS compared to healthy controls, ?Scan is the time interval between baseline and follow-up, and ? is the mixed-effects term accounting for subject-specific variability. The Bonferroni method was used to correct for multiple comparisons, and p-value <0.005 was considered significant.
Results: A total of 17 patients (n=8 WBRT; n=9 SRS) and 980 healthy controls were analyzed in the study. The median age was 61 years, 64 years, and 71 years for the WBRT, SRS and healthy control groups, respectively. 25% of WBRT patients and 22% of SRS patients had five or more brain metastases. Median (range) dose was 30 Gy (30-35 Gy) for WBRT and 20 Gy (16-21 Gy) for SRS. Median follow-up was 4.3 months and 6.0 months for WBRT and SRS groups respectively. The whole brain aged 10.93 times faster with WBRT (p<0.0001) and 8.077 times faster with SRS (p<0.0001) compared to controls. Similar aging pace was observed in the cortex for WBRT and SRS groups (f=10.97 WBRT, p<0.0001; f=7.91 SRS, p<0.0001). However, compared to controls, only WBRT but not SRS was associated with accelerated aging in the hippocampus (f=11.43 WBRT, p<0.0001; f=1.11 SRS, p=0.57) and all subcortical structures combined (f=11.97 WBRT, p<0.0001; f=0.75 SRS, p=0.74).
Conclusion: Compared to healthy controls, patients who received brain RT underwent structural changes resembling rapid aging. Accelerated aging pace was similar between patients treated with WBRT and SRS in terms of the whole brain and the cortex. However, WBRT demonstrated accelerated aging of subcortical structures including hippocampus while SRS did not; this mirrors the observed differences in neurocognitive outcomes between WBRT and SRS seen in prior studies. Further analysis with larger WBRT and SRS cohorts is ongoing.