NIMG-73. PREDICTING PERITUMORAL GLIOBLASTOMA INFILTRATION USING QUANTITATIVE MAGNETIC RESONANCE FINGERPRINTING RADIOMICS

Abstract PURPOSE Standard clinical magnetic resonance (MR) imaging of glioblastoma (GB) cannot identify malignant infiltration into the peritumoral non-enhancing region. This work implements quantitative MR fingerprinting (MRF) based radiomics to predict infiltrative regions GB. METHODS Pre-operative MRF T1 and T2 maps along with multiparametric (mpMR) images (T1w, T2w, T1w-Gd, FLAIR, ADC) from GB patients (n = 10) were analyzed. All subjects had histologically confirmed sites as identified by intra-operative 5-ALA fluorescence-guided tissue resection. These locations 40) manually annotated on pre-operative FLAIR a board-certified neuroradiologist labeled “infiltration” (INF). For each patient, another region ( > 3 cm enhancing tumor margin) was “edema” (ED). Following image co-registration, 693 handcrafted radiomic voxel-based features extracted (INF) edema (ED) voxels using 3D 5x5x5 voxel sliding kernel. Feature selection performed minimum redundancy maximal relevance (MRMR) algorithm, then INF ED used train two cross-validated binary support vector machine (SVM) models for voxel-wise prediction: 1) only model 2) combined mpMRI model. MRMR selection, included T1, T1w, ADC images. Model performance evaluated one withheld subject multiple known sites. Balanced test accuracy receiver operating characteristic (ROC) curve analysis evaluate classification performance. RESULTS The achieved balanced accuracies 77.87% 73.21% AUCs 0.91 0.97, respectively. CONCLUSIONS study demonstrates that MRF-based can high provide complementary value standard MRI features. Our results indicate potential employing guide personalized treatment strategies.