486 Surface-based deep learning model assessing brain aging after intracranial radiation for brain metastases

Objectives/Goals: Cognitive decline is a known sequalae of intracranial radiation in the treatment of brain metastases. In this study, we investigate global structural changes in the brain akin to accelerated aging and compare aging kinetics between patients treated with whole-brain radiation therapy (WBRT) and stereotactic radiosurgery (SRS). Methods/Study Population: This retrospective study consists of patients with brain metastases treated with WBRT and SRS at our institution. Brain MRI images collected prior to radiation therapy and at approximately three and six months following radiation will be analyzed, excluding patients with evidence of worsening disease burden in the brain. Surface morphology of the cerebral cortex and sub-cortical structures will be extracted using Freesurfer and converted to graphs. Data will then be input into a validated graph convolutional neural network model to estimate brain age at each time point. A generalized linear model will be used to estimate the aging pace between baseline and follow-up for each subject within the whole brain as well as the sub-cortical structures, which will be compared between WBRT and SRS treatment groups. Results/Anticipated Results: We anticipate that intracranial radiation will accelerate brain aging to a greater extent following WBRT compared to SRS. Additionally, this accelerated aging will occur globally in the whole brain as well as within individual substructures, including the cerebral cortex, nucleus accumbens, amygdala, caudate, hippocampus, pallidum, putamen, and thalamus. Discussion/Significance of Impact: This study will demonstrate structural changes in the brain analogous to accelerated aging, supporting its potential use as an imaging biomarker to monitor cognitive decline after radiation therapy. Future work will explore the relationship between structural brain aging and assessments of neurocognitive function.