Purpose: Advanced MRI techniques may provide non-invasive insight into the molecular heterogeneity of glioblastoma. Amide proton transfer-weighted (APTw) chemical exchange saturation transfer (CEST) MRI reflects endogenous protein and peptide content, but its clinical and molecular correlates in therapy-naive glioblastoma, IDH-wildtype, remain incompletely understood.
Methods: This retrospective single-center study included 53 adult patients with therapy-naive glioblastoma, IDH-wildtype, who underwent preoperative APTw MRI. Median time between imaging and tissue sampling was two days. Median and 90th percentile (p90) APTw signal intensities were extracted from contrast-enhancing (T1-CE) tumor regions and FLAIR-hyperintense regions using automated deep learning-based segmentation with manual quality control. Histological and molecular analyses included MGMT promoter methylation, Ki-67 index, and DNA methylation-based subclassification. Associations were assessed using non-parametric tests, multivariable linear regression, and Cox regression analyses.
Results: APTw signal intensity was significantly higher in T1-CE tumor regions than in FLAIR-hyperintense regions (p < 0.0001). Within the T1-CE region, higher APTw signal intensity was modestly associated with younger age. Glioblastomas of the mesenchymal methylation subtype demonstrated significantly higher median and p90 APTw signal intensity compared with RTK1 and RTK2 subtypes, independent of MGMT status and Ki-67 index. APTw signal intensity was not independently associated with PFS or OS.
Conclusion: APTw CEST MRI reflects molecular heterogeneity in therapy-naive IDH-wildtype glioblastoma, with potentially increased signal intensity in the mesenchymal subtype. These findings support its possible role as a complementary imaging biomarker for non-invasive molecular characterization.
Keywords: Amide proton transfer-weighted imaging; Chemical exchange saturation transfer imaging; DNA methylation; Glioblastoma; MRI; Mesenchymal subtype glioblastoma; Receptor tyrosine kinase subtype glioblastoma.
© 2026. The Author(s).