Potential use of iTB Maps to Distinguish GBM Recurrence from Pseudoprogression Based on Performance in the Context of Tumor Infiltration

Motivation: The motivation for this study is to address the unmet need to distinguish glioblastoma recurrence from pseudoprogression (9009) Goal(s): The goal is to assess the utility of a machine learning approach for distinguishing tumor infiltration from peritumoral edema in the context of solution features outlined in the 2025 challenge. Approach: Multiparametric MRI (mpMRI) and biopsies from glioma were used as inputs to the infiltrative tumor burden (iTB) model to predict the presence of tumor within non-contrast enhancing, FLAIR enhancing regions, and subsequent iTB maps were generated. Results: Performance and repeatability metrics validate this as a potentially useful approach for the unmet need. Impact: The validation of iTB maps the context of tumor infiltration and in terms of the unmet need required features indicate that a similar approach may be utilized to distinguish glioblastoma recurrence from pseudoprogression.