Abstract
Purpose To solve the challenge of interpreting diffusion tensor imaging (DTI), we proposed a joint histogram analysis of the isotropic (p) and anisotropic (q) components of DTI. We explored the heterogeneity of glioblastoma infiltration using the joint histogram features and evaluated their prognostic values.
Materials and methods A total of 115 primary glioblastoma patients were prospectively recruited and preoperatively imaged. Patients underwent maximal safe resection. DTI was processed and decomposed into p and q components. Pixel values were extracted from DTI-p and-q maps and used to construct the univariate and joint histograms, in contrast-enhancing and non-enhancing regions respectively. Eight joint histogram features were obtained and then correlated with patient survival and tumor progression rate. Their prognostic values were examined and compared with clinical factors using receiver operating characteristic curves.
Results Both univariate and joint histogram showed that the subregion of increased DTI-p and decreased DTI-q accounted for the largest proportion. However, additional diffusion patterns can be identified via joint histogram analysis. Particularly, a higher proportion of decreased DTI-p and increased DTI-q in non-enhancing region contributed to worse progression-free survival and worse overall survival (both HR = 1.12, p < 0.001); the proportion of this subregion showed a positive correlation (p = 0.010, r = 0.35) with tumor progression rate.
Conclusion Joint histogram analysis of DTI can provide a comprehensive measure of heterogeneity in infiltration and prognostic values for glioblastoma patients. The subregion of decreased DTI-p and increased DTI-q in non-enhancing region may indicate a more invasive habitat.
Funding This study was funded by a National Institute for Health Research (NIHR) Clinician Scientist Fellowship (SJP, project reference NIHR/CS/009/011); CRUK core grant C14303/A17197 and A19274 (FM lab); Cambridge Trust and China Scholarship Council (CL & SW); the Chang Gung Medical Foundation and Chang Gung Memorial Hospital, Keelung, Taiwan (JLY); CRUK & EPSRC Cancer Imaging Centre in Cambridge & Manchester (FM & TT, grant C197/A16465); Royal College of Surgeons of England (RS); NIHR Cambridge Biomedical Research Centre (TM & SJP). The Human Research Tissue Bank is supported by the NIHR Cambridge Biomedical Research Centre. We would like to acknowledge the support of National Institute for Health Research, the University of Cambridge, Cancer Research UK and Hutchison Whampoa Limited. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.
Conflict of Interest none
Manuscript type Original research
Advances in knowledge
Joint histogram analysis of the isotropic (p) and anisotropic (q) components of the diffusion tensor imaging can reflect the intratumoral heterogeneity of glioblastoma infiltration.
Incremental prognostic values for the prediction of overall survival and progression-free survival can be achieved by the joint histogram features, when integrated with IDH-1 mutation, MGMT methylation status and other clinical factors.
The non-enhancing tumor subregion in which water molecules display decreased isotropic movement and increased anisotropic movement are potentially representative of a more invasive tumor habitat.
Implications for patient care This study helps us to understand how the infiltrative patterns of glioblastoma contribute to patient outcomes. The invasive subregion identified by this approach may have clinical implications for personalized surgical resection and targeted radiation therapy.
Summary Statement Although diffusion tensor imaging has been accepted as a sensitive method in depicting white matter disruption, the difficulty in interpreting high-dimensional tensor may hinder its further clinical application. Joint histogram analysis of the decomposed components of tensor imaging is proposed as a potential solution for this challenge, and a reference to other imaging analysis using multiple biomarkers.