[Paper Review] The Rio Hortega University Hospital Glioblastoma dataset: a comprehensive collection of preoperative, early postoperative and recurrence MRI scans (RHUH-GBM)

Santiago Cepeda, Sergio García-García+5 more|PubMed|Apr 27, 2023

Glioma Diagnosis and Treatment13 references4 citations

TL;DR

The RHUH-GBM dataset provides a comprehensive collection of multiparametric preoperative, early postoperative (within 72 hours), and recurrence MRI scans from 40 glioblastoma patients who underwent gross total or near-total resection. It includes expert-corrected tumor subregion segmentations, clinical data, molecular profiles, and survival outcomes, enabling advanced research in postoperative MRI analysis, recurrence modeling, and AI-driven segmentation and registration algorithms.

ABSTRACT

Glioblastoma, a highly aggressive primary brain tumor, is associated with poor patient outcomes. Although magnetic resonance imaging (MRI) plays a critical role in diagnosing, characterizing, and forecasting glioblastoma progression, public MRI repositories present significant drawbacks, including insufficient postoperative and follow-up studies as well as expert tumor segmentations. To address these issues, we present the "Río Hortega University Hospital Glioblastoma Dataset (RHUH-GBM)," a collection of multiparametric MRI images, volumetric assessments, molecular data, and survival details for glioblastoma patients who underwent total or near-total enhancing tumor resection. The dataset features expert-corrected segmentations of tumor subregions, offering valuable ground truth data for developing algorithms for postoperative and follow-up MRI scans.

Motivation & Objective

• To address the lack of public MRI datasets with early postoperative and recurrence scans for glioblastoma patients.
• To provide high-quality, expert-verified tumor subregion segmentations for preoperative, early postoperative, and recurrence time points.
• To support the development of AI-based registration and segmentation algorithms tailored for postoperative and follow-up MRI scans.
• To enable research into recurrence patterns and prognostic modeling in patients with gross total or near-total resection.
• To enhance the availability of multimodal MRI, clinical, and molecular data for glioblastoma research.

Proposed method

• Collected multiparametric MRI scans (T1w, T1ce, T2w, FLAIR, ADC) at three time points: preoperative, early postoperative (≤72 h), and recurrence.
• Selected 40 patients with WHO grade 4 astrocytoma who underwent gross total resection (GTR) or near-total resection (NTR) >95%.
• Performed image preprocessing using dcm2niix for DICOM-to-NIfTI conversion, FLIRT for linear registration to the SRI24 atlas, and Synthstrip for skull-stripping.
• Applied intensity Z-scoring normalization to standardize MRI signal intensities across subjects.
• Generated expert-corrected segmentations of enhancing tumor, non-enhancing tumor, and peritumoral T2/FLAIR hyperintensities at each time point.
• Integrated clinical, pathological, and treatment data including KPS, IDH status, Stupp protocol therapy, and survival outcomes.

Experimental results

Research questions

• RQ1 How do tumor subregions evolve across preoperative, early postoperative, and recurrence MRI scans in glioblastoma patients with gross total resection?
• RQ2 Can expert-annotated segmentations from early postoperative scans improve the accuracy of AI-based tumor segmentation and registration algorithms?
• RQ3 What are the patterns of recurrence and peritumoral signal changes in patients with high extent of resection, as captured by multiparametric MRI?
• RQ4 How do molecular markers (e.g., IDH status) and clinical factors correlate with recurrence timing and survival in this cohort?
• RQ5 To what extent can the RHUH-GBM dataset improve the development of prognostic models for glioblastoma using multimodal MRI and clinical data?

Key findings

• The dataset comprises 40 glioblastoma patients with complete preoperative, early postoperative (within 72 hours), and recurrence MRI scans, all acquired at the same institution.
• All patients underwent gross total resection (GTR) or near-total resection (NTR) with >95% resection of the enhancing tumor volume.
• The median overall survival was 364 days, and the median progression-free survival was 198 days.
• The dataset includes expert-corrected segmentations of enhancing tumor, non-enhancing tumor, and peritumoral T2/FLAIR hyperintensities at all three time points.
• All patients received adjuvant chemotherapy and radiotherapy according to the Stupp protocol, with 72.5% receiving 60 Gy in 30 fractions via VMAT-IMRT-IGRT.
• The dataset is publicly available via The Cancer Imaging Archive (TCIA) and includes preprocessing code on GitHub for reproducibility.