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.
Keywords: MR imaging, Neuro-oncology, Radiology, Brain tumor
Specifications Table
| Subject | Health and Medical Sciences:
|
| Specific subject area | Magnetic resonance imaging (MRI) data and tumor segmentations from patients with glioblastoma. |
| Data format | Raw |
| Type of data | The dataset contains MRI data including the sequences: T1-weighted (T1w), T2-weighted (T2w), fluid attenuated inversion recovery (FLAIR), T1w contrast-enhanced (T1ce), and diffusion-weighted imaging-derived apparent diffusion coefficient (ADC) maps. Data is available in both raw DICOM and processed NIfTI format along with expertly refined tumor segmentations. Clinical information is also available in CSV format. |
| Data collection | MRI data were collected retrospectively from the hospital's picture archiving and communication system (PACS). Imaging data were acquired on a 1.5 Tesla MRI scanner and consists of multiparametric structural and diffusion MRI images acquired at three time points: preoperatively, early, and at follow-up when tumor recurrence was diagnosed. In addition, the dataset includes only patients who underwent total or near-total resection of the enhancing tumor. Clinical data were collected from electronic medical records. Tumor subregion segmentations were generated by computer aided methods and carefully reviewed and manually corrected by two expert neurosurgeons specializing in neuroimaging. |
| Data source location | The Río Hortega University Hospital, Valladolid, Spain. |
| Data accessibility | Repository name: The Cancer Imaging Archive Data identification number: https://doi.org/10.7937/4545-c905 Direct URL to data: https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=145755234 Instructions for accessing these data: Raw DICOM data in this collection contains images that could potentially be used to reconstruct a human face. To safeguard the privacy of participants, users must sign and submit a TCIA Restricted License Agreement |
1. Value of the Data
-
•
The value of The Río Hortega University Hospital Glioblastoma Dataset (RHUH-GBM) [1] lies in the inclusion of longitudinal MRI scans obtained at critical points in the disease course: pre-surgery, early post-surgery, and at the time of recurrence. It is important to note that patients in this cohort underwent either gross total or near-total tumor resection, further enhancing the dataset's significance. Moreover, the dataset features meticulously refined expert segmentations, contributing to its overall richness.
-
•
Researchers and medical professionals alike can leverage this dataset for a wide range of research purposes. These applications include enhancing automatic segmentation algorithms tailored to brain tumor postoperative scans, developing models for predicting survival rates, and investigating recurrence patterns in patients who have undergone complete tumor resection.
-
•
Furthermore, the dataset is accessible in two distinct formats: NiFTI and DICOM, and it is readily available on the TCIA website. Additionally, the accompanying clinical data is comprehensive, encompassing demographic, pathological, radiological, volumetric, and survival information, further enriching its utility for research endeavors.
2. Data Description
2.1. Shared files
Table 1 shows the details of the files available through TCIA. DICOM and NiFTI files can be visualized in dedicated and publicly available software such as 3D Silicer (www.slicer.org) and ITK-SNAP (www.itksnap.org). Clinical data is also available in CSV format.
Table 1.
Detailed dataset description.
| Number of Patients | 40 |
| Number of Studies | 120 |
| Number of Series | 600 |
| Number of Raw Images | 37,425 (DICOM) |
| Number of Processed Images + Tumor segmentations |
720 (NIfTI) |
| Images Size (GB) | 16 GB (DICOM) + 2.9 GB (NIfTI) |
2.2. Patient population
The dsataset comprises consecutive patients who underwent surgery between January 2018 and December 2022, with a confirmed histopathological diagnosis of WHO grade 4 astrocytoma. Forty patients were selected based on the following inclusion criteria: 1) Gross total resections (GTR) or Near Total Resection (NTR), defined as having no residual tumor enhancement and an extent of resection exceeding 95% of the initial enhancing volume, respectively [2,3]. 2) Availability of MRI studies at three time points: preoperative, early postoperative (within 72 h), and the follow-up scan where tumor progression was diagnosed. 3) Availability of structural T1-weighted (T1w), T2-weighted (T2w), T1 contrast-enhanced (T1ce), Fluid-attenuated inversion recovery (FLAIR), and diffusion-weighted imaging-derived apparent diffusion coefficient (ADC) maps for each study. 4) Receipt of adjuvant treatment with chemotherapy and radiotherapy following the Stupp protocol [4]. Patients with severe image acquisition artifacts or missing MRI series were excluded. The modified Response Assessment in Neuro-Oncology (RANO) criteria were utilized to determine tumor progression [5].
A summary of the demographic data is presented in Table 2. The patients had an average age of 63 ± 9 years, consisting of 28 men (70%) and 12 women (30%). The median preoperative Karnofsky Performance Scale (KPS) score was 80. Out of the 40 patients, 38 (95%) were diagnosed with de novo glioblastomas, while two patients (5%) had recurrent glioblastomas previously treated with standard chemoradiotherapy. Four cases (10%) were IDH-mutated, and 36 cases (90%) were IDH wild-type.
Table 2.
Study population demographics of the Río Hortega University Hospital Glioblastoma dataset (RHUH-GBM).
| Sex | Male | 28 (70%) |
| Female | 12 (30%) | |
| Age (years) | 63 ± 9 | |
| Extent of resection | GTR | 27 (67.5%) |
| NTR | 13 (32.5%) | |
| Number of time-point MRI studies | 120 | |
| Number of MRI series | 600 | |
| IDH status | Mutant | 4 (10%) |
| Wild type | 36 (90%) | |
| Preoperative KPS | 80 (10) | |
| Operative adjuncts | 5’ALA | 40 (100%) |
| Sodium Fluorescein | 7 (17.5%) | |
| Neuronavigation | 40 (100%) | |
| ioUS | 40 (100%) | |
| IONM | 4 (10%) | |
| DES | 3 (7.5%) | |
| Preoperative contrast enhancing tumor volume (cm3) | 34.99 ± 26.59 | |
| Preoperative T2/FLAIR peritumoral signal alteration volume (cm3) | 35.00 ± 26.74 | |
| Postoperative contrast enhancing tumor volume (cm3) | 0.23 ± 0.47 | |
| Postoperative T2/FLAIR peritumoral signal alteration volume (cm3) | 35.00 ± 26.74 | |
| Radiotherapy treatment details | VMAT-IMRT-IGRT/60 Gy /30 fx | 29 (72.5%) |
| VMAT-IMRT-IGRT/50 Gy /20 fx | 6 (15%) | |
| VMAT-IMRT-IGRT/40.5 Gy /15 fx | 5 (12.5%) | |
| Postoperative neurological deficit | No | 26 (65%) |
| Transient | 6 (15%) | |
| Minor persistent | 6 (15%) | |
| Major persistent | 2 (5%) | |
| Postoperative KPS | 80 (20) | |
Numerical values are expressed in mean and standard deviation or median and interquartile range accordingly. GTR = gross total resection, NTR = near total resection, IDH = isocitrate dehydrogenase, 5’ALA = 5-aminolevulinic acid, ioUS = intraoperative ultrasound, IONM = Intraoperative Neurophysiological Monitoring, DES = direct electrical stimulation, VMAT = Volumetric Modulated Arc Therapy, IMRT = Intensity Modulated Radiation Therapy, IGRT = Image-Guided Radiation Therapy. Radiotherapy treatments are expressed in dose (Gy) and number of fractions (fx).
The mean preoperative contrast-enhancing tumor volume was 34.99 ± 26.59 cm3, and the mean postoperative contrast-enhancing residual tumor volume was 0.23 ± 0.47 cm3. A graphical representation of tumor location is displayed as a heatmap in Fig. 1. Among the patients, 27 (67.5%) underwent gross total resection, and 13 (32.5%) underwent near-total resection. The median overall survival was 364 days, and the median progression-free survival was 198 days.
Fig. 1.
A graphical representation of tumor location is presented in the form of a heatmap, showcasing the distribution of tumors in a normalized SRI24 atlas template space. Areas of interest are depicted as percentages.
2.3. Clinical, pathological, and imaging data
Clinical and pathological information was obtained from electronic medical records, including age, sex, histopathological diagnosis, pre- and postoperative Karnofsky Performance Score (KPS), isocitrate dehydrogenase (IDH) status, use of operative adjuncts, volumetric assessment of the extent of resection of the contrast-enhancing and non-enhancing tumor, presence of postoperative neurological deficits, details of chemotherapy and radiotherapy received, and overall survival (OS) and progression-free survival (PFS) times. OS was measured from diagnosis to death or last follow-up if alive, while PFS was from diagnosis to tumor progression or last follow-up if no progression was noted. Out of the total sample, a subset of 11 patients had initially undergone preoperative and subsequent follow-up MRI scans at a secondary healthcare facility before being referred to the primary center. Details of the MR imaging acquisition parameters are described in Table 3.
Table 3.
MRI acquisition parameters.
| Primary center |
Secondary center |
|||
|---|---|---|---|---|
| Manufacturer, model, and Field strength |
General electric, signa HDxT, 1.5 T |
Philips, ingenia ambition X, 1.5 T |
Philips, achieva, 1.5 T |
|
| Number of MRI studies | 107 (89 %) | 11 (9 %) | 2 (2 %) | |
| MRI sequence | T1ce | TR/TE, 7.98 ms/2.57 ms; 3D; GRE; matrix, 512 × 512; slice thickness, 1 mm | TR/TE, 17.96 ms/6.43 ms; 3D ProSET; matrix, 230 × 230; slice thickness, 1 mm | TR/TE, 25 ms/6.7 ms; ProSET, 3D; matrix, 256 × 256; slice thickness, 1.6 mm |
| T1w | TR/TE, 580 ms/7.56 ms; 2D; FSE; matrix, 512 × 512; slice thickness, 5 mm | TR/TE, 525.6 ms/12 ms; 2D; SE; matrix, 228 × 227; slice thickness, 5 mm | TR/TE, 456.2 ms/12 ms; 2D; SE; matrix, 249 × 191; slice thickness, 6 mm | |
| T2w | TR/TE, 5220 ms/96.12 ms; 2D; FRSE; matrix, 512 × 512; slice thickness, 5 mm. | TR/TE, 5327.3 ms/110 ms; 2D; TSE; matrix, 232 × 232; slice thickness, 3 mm. | TR/TE, 2456.2 ms/110 ms; 2D; TSE; matrix, 264 × 203; slice thickness, 5 mm. | |
| FLAIR | TR/TE, 8002 ms/135.07 ms; 2D; FSE; matrix, 512 × 512; slice thickness, 4 mm | TR/TE, 5000 ms/375.8 ms; 3D; SPIR; matrix, 196 × 196; slice thickness, 1.2 mm | TR/TE, 6000 ms/120 ms; 2D; FSE; matrix, 200 × 159; slice thickness, 2.8 mm | |
| DWI | TR/TE, 8000 ms/111.7 ms; matrix, 128 × 160; slice thickness, 5 mm; b-values, 0 and 1000 s/mm2 | TR/TE, 4600 ms/84.4 ms; matrix, 190 × 190; slice thickness, 5 mm; b-values, 0 and 1000 s/mm2 | TR/TE, 3414 ms/88.8 ms; matrix, 112 × 89; slice thickness, 5 mm; b-values, 0 and 1000 s/mm2 | |
T1ce = contrast-enhanced T1w, T2w= T2-weighted image, FLAIR = Fluid-attenuated inversion recovery, DWI = diffusion weighted image, TR = repetition time, TE= echo time, GRE = gradient echo. FSE= fast spin echo. FRFSE= fast recovery fast spin echo. ProSET= principle of selective excitation technique. SE= spin echo. SPIR= spectral presaturation with inversion recovery. TSE= turbo spin echo.
3. Experimental Design, Materials and Methods
3.1. Image preprocessing
Images were retrieved from the Picture Archiving Communication System (PACS) in Digital Imaging and Communications in Medicine (DICOM) format for subsequent processing. The first step involved converting the images to Neuroimaging Informatics Technology Initiative (NIfTI) format using the dicom2niix tool version v1.0.20220720 [6], available at https://github.com/rordenlab/dcm2niix/releases/tag/v1.0.20220720. Subsequently, the T1ce scans for each subject were registered to the SRI24 anatomical atlas space [7] using the FLIRT (FMRIB's Linear Image Registration Tool) [8,9] available at https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSL. The T1w, T2w, FLAIR scans, and ADC maps were then registered to the transformed T1ce scan, resulting in co-registered resampled volumes of 1 × 1 × 1 mm isotropic voxels. The brain was extracted from all co-registered scans using a deep learning tool called Synthstrip [10], included in FreeSurfer v7.3.0, available at https://github.com/freesurfer/freesurfer/tree/dev/mri_synthstrip. Finally, intensity Z-scoring normalization was performed for structural MRI sequences (T1ce, T2w, T1w, and FLAIR) using the normalization tools included in Cancer Imaging Phenomics Toolkit (CaPTk) v1.9.0 [11] available at https://www.nitrc.org/projects/captk/. Intensity normalization was not performed on ADC maps since intensity values are already standardized and expressed in units of mm2/s. The code applied for image preprocessing is available to the public through a GitHub repository https://github.com/smcch/RHUH-GBM-dataset-MRI-preprocessing.
3.2. Tumor Subregions Segmentations
The preprocessed images from each time point were used as input for generating computer-aided segmentations using Deep-Medic [12]. Three labels were subsequently obtained, corresponding to 1 - necrosis, 2 - peritumoral signal alteration, including edema and non-enhancing tumor, and 3 - enhancing tumor. All segmentations were carefully reviewed and manually corrected by two expert neurosurgeons specializing in neuroimaging (S.C. and S.G.). A summary of the data workflow is depicted in Fig. 2.
Fig. 2.
Schematic representation of the workflow. It displays the main clinical variables collected, the image preprocessing steps, and the results of the tumor subregion segmentations: red = necrosis, green = peritumoral region (T2/FLAIR) signal alteration, yellow = enhancing tumor.
Limitations
Not applicable.
Ethics Statement
The study was conducted following the principles of the Declaration of Helsinki. Written consent was obtained from all patients, and approvals were granted by the Institutional Review Board of Río Hortega University Hospital and the Ethics Committee for Drug Research (CEIm) of the West Valladolid Health Area (Ref. 22PI-208).
CRediT authorship contribution statement
Santiago Cepeda: Conceptualization, Methodology, Software, Funding acquisition, Data curation, Writing – original draft. Sergio García-García: Data curation, Methodology, Software, Writing – review & editing. Ignacio Arrese: . Francisco Herrero: Resources, Visualization, Investigation. Trinidad Escudero: Resources, Visualization, Investigation. Tomás Zamora: Resources, Visualization, Investigation. Rosario Sarabia: .
Acknowledgements
This work was partially funded by a grant awarded by the “Instituto Carlos III, Proyectos I-D-i, Acción Estratégica en Salud 2022”, under the project titled “Prediction of tumor recurrence in glioblastomas using magnetic resonance imaging, machine learning, and transcriptomic analysis: A supratotal resection guided by artificial intelligence,” reference PI22/01680.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Data Availability
RHUH-GBM (Original data) (https://www.cancerimagingarchive.net/).
References
- 1.S. Cepeda, S. García García, I. Arrese, F. Herrero, T. Escudero Caro, T. Zamora Martínez, R. Sarabia, The Río Hortega University Hospital Glioblastoma dataset: a comprehensive collection of preoperative, early postoperative and recurrence MRI scans (RHUH-GBM), (2023). 10.7937/4545-C905. [DOI] [PMC free article] [PubMed]
- 2.Karschnia P., Young J.S., Dono A., Häni L., Sciortino T., Bruno F., Juenger S.T., Teske N., Morshed R.A., Haddad A.F., Zhang Y., Stoecklein S., Weller M., Vogelbaum M.A., Beck J., Tandon N., Hervey-Jumper S., Molinaro A.M., Rudà R., Bello L., Schnell O., Esquenazi Y., Ruge M.I., Grau S.J., Berger M.S., Chang S.M., van den Bent M., Tonn J.-C. Prognostic validation of a new classification system for extent of resection in glioblastoma: a report of the RANO resect group. Neuro-Oncol. 2022 doi: 10.1093/neuonc/noac193. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Karschnia P., Vogelbaum M.A., van den Bent M., Cahill D.P., Bello L., Narita Y., Berger M.S., Weller M., Tonn J.-C. Evidence-based recommendations on categories for extent of resection in diffuse glioma. Eur. J. Cancer (Oxford, England : 1990) 2021;149:23–33. doi: 10.1016/j.ejca.2021.03.002. [DOI] [PubMed] [Google Scholar]
- 4.Stupp R., Hegi M.E., Mason W.P., van den Bent M.J., Taphoorn M.J.B., Janzer R.C., Ludwin S.K., Allgeier A., Fisher B., Belanger K., Hau P., Brandes A.A., Gijtenbeek J., Marosi C., Vecht C.J., Mokhtari K., Wesseling P., Villa S., Eisenhauer E., Gorlia T., Weller M., Lacombe D., Cairncross J.G., Mirimanoff R.-O. European organisation for research and treatment of cancer brain tumour and radiation oncology groups, National Cancer Institute of Canada Clinical Trials Group, effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet. Oncol. 2009;10:459–466. doi: 10.1016/S1470-2045(09)70025-7. [DOI] [PubMed] [Google Scholar]
- 5.Ellingson B.M., Wen P.Y., Cloughesy T.F. Modified criteria for radiographic response assessment in glioblastoma clinical trials. Neurotherapeutics. 2017;14:307–320. doi: 10.1007/s13311-016-0507-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Li X., Morgan P.S., Ashburner J., Smith J., Rorden C. The first step for neuroimaging data analysis: DICOM to NIfTI conversion. J. Neurosci. Methods. 2016;264:47–56. doi: 10.1016/j.jneumeth.2016.03.001. [DOI] [PubMed] [Google Scholar]
- 7.Rohlfing T., Zahr N.M., Sullivan E.V, Pfefferbaum A. The SRI24 multichannel atlas of normal adult human brain structure. Hum. Brain Mapp. 2010;31:798–819. doi: 10.1002/hbm.20906. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Jenkinson M., Smith S. A global optimisation method for robust affine registration of brain images. Med. Image Anal. 2001;5:143–156. doi: 10.1016/s1361-8415(01)00036-6. [DOI] [PubMed] [Google Scholar]
- 9.Jenkinson M., Bannister P., Brady M., Smith S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage. 2002;17:825–841. doi: 10.1016/s1053-8119(02)91132-8. [DOI] [PubMed] [Google Scholar]
- 10.Hoopes A., Mora J.S., Dalca A.V, Fischl B., Hoffmann M. SynthStrip: skull-stripping for any brain image. Neuroimage. 2022;260 doi: 10.1016/j.neuroimage.2022.119474. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Davatzikos C., Rathore S., Bakas S., Pati S., Bergman M., Kalarot R., Sridharan P., Gastounioti A., Jahani N., Cohen E., Akbari H., Tunc B., Doshi J., Parker D., Hsieh M., Sotiras A., Li H., Ou Y., Doot R.K., Bilello M., Fan Y., Shinohara R.T., Yushkevich P., Verma R., Kontos D. Cancer imaging phenomics toolkit: quantitative imaging analytics for precision diagnostics and predictive modeling of clinical outcome. J. Med. Imag. (Bellingham, Wash.) 2018;5 doi: 10.1117/1.JMI.5.1.011018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kamnitsas K., Ledig C., Newcombe V.F.J., Simpson J.P., Kane A.D., Menon D.K., Rueckert D., Glocker B. Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation. Med. Image Anal. 2017;36:61–78. doi: 10.1016/j.media.2016.10.004. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
RHUH-GBM (Original data) (https://www.cancerimagingarchive.net/).