NIRA-04 DIFFERENTIATION OF BRAIN TUMOR PROGRESSION AND TREATMENT EFFECT USING DIFFUSION-RELAXATION CORRELATION SPECTROSCOPIC IMAGING

Abstract Differentiating treatment effects (TxE) from tumor progression in glioma and metastatic brain tumors is challenging due to their similar clinical presentations and conventional MRI characteristics. Scenarios with mixed pathology further complicate diagnosis. Diffusion-Relaxation Correlation Spectroscopic Imaging (DR-CSI) is an emerging technique enabling microscopic tissue components characterization by measuring joint variations in T2 relaxation and diffusivity. This study aims to evaluate the feasibility of DR-CSI for differentiating TxE from viable tumors. Thirteen patients with suspected brain tumors (glioma or brain metastasis) and/or suspected TxE underwent preoperative MRI, including DR-CSI and a standard clinical protocol with diffusion and perfusion sequences, on 3T MRI systems. DR-CSI data were acquired with a 30 TE-b-value encoding diffusion-weighted spin-echo EPI sequence. Histopathology confirmed 6 tumor cases, 4 pure TxE cases, and 3 with mixed pathology. Voxel-wise diffusion-relaxation (D-T2) spectra were calculated using non-negative least-squares fitting with total variation regularization. Regions of interest (ROIs) encompassing contrast-enhancing lesions and necrotic areas were segmented using NEOSOMA. D-T2 spectra were extracted from lesion ROIs, and Mann-Whitney U tests were applied to compare spectral features between tumor and TxE groups. Spectral regions showing significant group differences were identified via U statistics. A “TxE fraction” was calculated by integrating the spectral power within the discriminative region. In TxE, mixed, and tumor groups, the TxE fractions were 0.14±0.04, 0.09±0.02, and 0.06±0.05, respectively. TxE fractions tended to differ across groups: tumor vs. non-tumor (p=0.05, AUC=0.83), TxE vs. non-TxE (p=0.05, AUC=0.83), and pure TxE vs. pure tumor (p=0.07, AUC=0.86, sensitivity=1.0, specificity=0.67). In contrast, conventional imaging biomarkers did not show significant differences across the groups: ADC (p=0.20, 0.84, 0.48) and rCBV (p=0.94, 0.73, 1.00). These preliminary results support the potential of DR-CSI as a non-invasive tool for resolving microstructural differences in brain lesions and improving differential diagnosis in the clinical setting.