Differentiating Recurrence from Pseudoprogression in High-Grade Gliomas Using Three-dimensional Pseudo-Continuous Arterial Spin Labeling

Motivation: Accurately distinguishing between postoperative recurrence(TR) and pseudoprogression(PsP) in glioma patients is crucial for guiding subsequent treatment strategies. Goal(s): This study explores the efficiency of 3D-pCASL technology in differentiating PsP and TR in postoperative glioma patients. Approach: Ninety postoperative glioma patients underwent MRI scanning, including 3D-pCASL. Cerebral blood flow (CBF) maps were generated, and then regions of interest were analyzed to calculate average values and extract radiomic features for building machine learning models. Results: Results showed the mean CBF values and models based on radiomic features extracted from CBF maps effectively distinguished PsP from TR, achieving AUC values of 0.860 and 0.841, respectively. Impact: 3D-pCASL and its radiomics can effectively identify TR versus PsP in postoperative glioma patients which is difficultly differentiated by conventional MRI sequences，Integrating machine learning to further optimize radiomic models could enhance diagnostic accuracy and improve patient survival rates.