Green Diversity Supplement: Predicting 5-ALA Fluorescence Status in High Grade Gliomas Based on MRI Features

Abstract High-grade gliomas (HGGs), such as glioblastoma multiforme (GBM), are notably aggressive, heterogeneous, and infiltrating brain tumors, presenting significant challenges in surgical resection and leading to poor survival rates. Despite advancements in surgical techniques, radiation, and chemotherapy, the invasive nature and high recurrence rate of HGGs limit

treatment effectiveness. Fluorescence-guided surgery with 5-aminolevulinic acid (5-ALA) offers high specificity and sensitivity in tumor margin delineation but is hindered by limitations such as false negatives due to photobleaching, obstruction by other tissues, low tumor cell density, and non-efficient dosage timing. These challenges can leave active tumor regions after surgery that

could lead to recurrence and complicate subsequent treatments. Additionally, T1-weighted imaging with gadolinium-based contrast (T1Gd) often underestimates the tumor burden, particularly in non-enhancing regions, leading to residual disease. Addressing these challenges, our study aims to identify MRI features correlating with 5-ALA positive and negative areas in

HGGs to develop a radiomics model predicting 5-ALA fluorescence on preoperative MRI scans. Our central hypothesis is that a radiomics model can predict 5-ALA fluorescence from MRI features in glioblastoma patients, and when considering sex differences, further refine its accuracy. With the proposed model, we intend to improve preoperative planning and surgical

outcomes by accurately identifying tumor margins. Furthermore, we will evaluate the model's prognostic utility by linking 5-ALA fluorescence predictions to the extent of tumor resection and survival rates. Given the emerging evidence of HGGs as a sexually dimorphic disease, our study will also explore sex differences in model development, anticipating significant impacts on

predictive accuracy and survival outcomes. The project aims to provide surgeons with objective evidence to assess tumor burden, plan surgeries more effectively, and improve survival outcomes across glioma patient groups. The project proposed here will be an extension of existing work by the Mathematical Neuro-Oncology (MNO) Lab (Parent Project PI: Dr. Kristin

Swanson), utilizing ongoing research in image-localized biopsies, MRI-based invasion mapping, and image-based model development.