IMG-127. Mulitmodal longitudinal analysis of topological features reveals differential changes in MRI in malignant glioma across species

Abstract INTRODUCTION One of the challenges of monitoring response to treatment in malignant glioma is detecting subtle changes in MR imaging that may indicate disease progression or therapeutic response. This is particularly difficult in the clinical trial setting, given small sample sizes, image registration variability, and heterogenous or novel radiographic morphology. Topological Data Analysis (TDA) and persistent homology can be used to quantify complex tumor shapes and heterogeneities from MRI data in this population. These tools provide robust geometric summaries that capture topological features like connected components and voids, enabling the identification of key radiological markers that may otherwise go undetected. METHODS We used persistent homology to extract topological features from T2 and FLAIR weighted MR images in the KR158 model of glioblastoma (GBM), pet dogs with canine glioma, and human GBM patients treated with RNA lipid particle aggregates (RNA-LPAs) (NCT06389591). These features were then compared to untreated controls. Results and Future Directions: There are clear differences in topological features of images from treated and untreated cohorts across species, despite small sample sizes and variability in image acquisition and registration. This data is rapidly translatable and applicable to monitoring of patients with malignant glioma undergoing treatment with novel therapeutics, particularly within immunotherapy clinical trials.