Pathological Validation of MRI Tumor Habitats in Glioblastoma

Background Introduction

Glioblastoma (GBM) is a highly malignant brain tumor characterized by significant heterogeneity and invasiveness. Due to its complex tumor microenvironment (TME), traditional imaging methods struggle to accurately distinguish different tumor regions, such as the tumor core, infiltrative tumor margins, and necrotic areas. This spatial heterogeneity not only affects treatment efficacy but also leads to poor patient prognosis. Therefore, accurately identifying different tumor regions through non-invasive imaging has become a focal point of current research.

In recent years, tumor habitat imaging based on multiparametric MRI (e.g., diffusion-weighted imaging, DWI, and dynamic susceptibility contrast-enhanced imaging, DSC) has gained traction. This method uses voxel-wise clustering analysis to identify distinct physiological regions within the tumor, providing deeper insights into the tumor’s biological characteristics. However, the correspondence between these imaging habitats and pathological features has not been fully validated. To address this, Ji Eun Park and colleagues conducted a study using data from the Ivy Glioblastoma Atlas Project (IvyGAP) to biologically validate MRI tumor habitats through pathology.

Source of the Study

The research was conducted by Ji Eun Park, Joo Young Oh, and other researchers from the Department of Radiology and Research Institute of Radiology at the University of Ulsan College of Medicine in Seoul, South Korea. The study was published online ahead of print on August 23, 2024, in Neuro-Oncology. The research team also included collaborators from multiple institutions, including the Department of Neurosurgery and the Department of Biochemistry and Molecular Biology at Asan Medical Center in Seoul.

Research Process and Results

1. Data Source and Patient Selection

The study utilized data from the IvyGAP project, which provides MRI images, pathological slides, and RNA sequencing data from 41 patients with IDH-wildtype glioblastoma. The research team selected data from 20 patients (22 tumors, 168 pathological slides) who had complete preoperative MRI scans (including T1, T2, FLAIR, DWI, and DSC imaging) and pathological data.

2. Construction of MRI Tumor Habitats

The research team first preprocessed the MRI images, including skull stripping, lesion segmentation, and image registration. Subsequently, using the apparent diffusion coefficient (ADC) maps generated from DWI and relative cerebral blood volume (rCBV) maps from DSC imaging, the tumor regions were divided into six habitats using a k-means clustering algorithm: - Contrast-Enhancing Lesion (CEL): Hypervascular habitat (C1), hypovascular cellular habitat (C2), and hypovascular hypocellular habitat (C3). - Non-Enhancing Lesion (NEL): Hypervascular habitat (C4), hypovascular cellular habitat (C5), and hypovascular hypocellular habitat (C6).

3. Processing of Pathological Data

Pathological slides were divided into multiple regions, including the leading edge (LE), infiltrating tumor (IT), cellular tumor (CT), hypervascular region (CThypervascular), and perinecrotic region (CTperinecrotic). The research team spatially registered these pathological regions with MRI images, calculated the normalized area of each pathological region, and performed correlation analysis with the voxel counts of MRI habitats.

4. Integration of RNA Sequencing Data

The study also utilized RNA sequencing data provided by IvyGAP to analyze the transcriptomic features of different pathological regions. Using the four glioblastoma subtypes proposed by Neftel et al. (mesenchymal-like, astrocyte-like, oligodendrocyte-progenitor-like, and neural-progenitor-like), the research team calculated the transcriptomic module scores for each pathological region and correlated them with MRI habitats.

5. Key Findings

• Correlation Between Pathology and MRI Habitats:

  • Cellular tumor (CT) showed a positive correlation with the hypovascular cellular habitat (C2) in CEL (r = 0.238, p = 0.005).
  • Infiltrating tumor (IT) showed a positive correlation with the hypovascular cellular habitat (C5) in NEL (r = 0.294, p = 0.017).
  • Hypervascular region (CThypervascular) showed a positive correlation with the hypervascular habitat (C4) in NEL (r = 0.195, p = 0.023).
  • Perinecrotic region (CTperinecrotic) showed a positive correlation with imaging necrosis (r = 0.199, p = 0.005).

• Correlation Between RNA Transcriptomics and Pathological Regions:

  • Astrocyte-like subtype showed a positive correlation with infiltrating tumor (IT) (r = 0.256, p < 0.001).
  • Mesenchymal-like subtype showed a positive correlation with the perinecrotic region (CTperinecrotic) (r = 0.246, p < 0.001).

Conclusions and Significance

The study biologically validated MRI tumor habitats through pathology, confirming the correspondence between the hypovascular cellular habitat in CEL and cellular tumor, as well as the hypovascular cellular habitat in NEL and infiltrating tumor. Additionally, the study revealed the transcriptomic characteristics of different tumor regions through RNA sequencing data, providing new insights into the heterogeneity of glioblastoma.

The scientific value of this research lies in: 1. Non-Invasive Tumor Monitoring: MRI tumor habitat analysis enables non-invasive identification of aggressive and infiltrative tumor regions, offering critical guidance for clinical treatment. 2. Treatment Guidance: The findings can guide surgical resection extent, radiotherapy target design, and early recurrence prediction. 3. Biological Validation: The integration of pathology and transcriptomic data provides strong evidence for the biological significance of imaging habitats.

Research Highlights

• Innovative Methodology: The study is the first to combine MRI tumor habitat analysis with pathology and transcriptomic data, offering a multidimensional analysis of tumor heterogeneity.
• Clinical Application Potential: The results provide new tools for personalized treatment of glioblastoma, particularly in surgery and radiotherapy.
• Data Accessibility: The research team has made all registered data and code publicly available, offering valuable resources for future studies.

Additional Valuable Information

The research team also highlighted the study’s limitations, such as potential errors introduced by manual registration of pathological slides and imaging, as well as the idealized settings of IvyGAP data that may differ from real-world clinical scenarios. Future studies should validate these findings in larger patient cohorts and explore the correspondence between more pathological regions and imaging habitats.

This study provides significant scientific evidence for the imaging diagnosis and treatment of glioblastoma, marking another important step in the clinical translation of tumor habitat analysis.