In the GLIOHAB project, a fully automated pipeline for processing MRI data was implemented to generate functional maps depicting vascular, diffusivity, as well as biomechanical characteristics of the lesion and surrounding tissue at the pixel level. This pipeline was used to process the project data, but also to analyze TCGA-GBM and Ivy Gap public datasets. The results of these analyses were published in open repositories for their re-use and enrichment of the original datasets [1,2]. In addition, this pipeline is currently being used within the NCT03951142 clinical trial at the Oslo University Hospital. Advances in the state-of-the-art and innovations related to the development of this pipeline were published in leading scientific journals and congresses [3–7] .
At GLIOHAB we have gone a step further in the characterization of intratumoral heterogeneity. From a more theoretical point of view, we developed a novel method of structural unsupervised segmentation that allows an improvement in the robustness of the heterogeneity characterization [8]. In addition, we developed a new concept of habitat linked to the characterization of tumor progression, which is currently under consideration for patenting.
Throughout the GLIOHAB project, we discovered relevant associations between vascular and mechanical habitats and the survival of glioblastoma patients. Moreover, in the case of vascular habitats we validated this relationship in an international multicenter setting [9]. We discovered of an association between vascularity in the peripheral infiltrated glioblastoma and different molecular subtypes of glioblastoma [10]. We assessed local microvascular proliferation in glioblastoma using relative Cerebral Blood Volume [11]. Finally, we discovered that the prognostic ability of vascularity in glioblastoma patients is higher in long-surviving patients [12].
Finally, in GLIOHAB we studied the clinical relevance of the characterization of tumor heterogeneity for treatment response. One of the most relevant findings is that combining MGMT methylation status together with vascularity estimated in the high angiogenic part of the tumor, obtained automatically from standard-of-care MRI studies, can provide a more reliable prognostic indicator for designing patient stratification strategies [13]. Another relevant finding is the association found between small peritumoral displacements due to tumor growth and survival throughout patient follow-up. The methodology of quantification of these displacements proposed by GLIOHAB could contribute significantly to the differentiation of pseudoprogression and tumor progression phenomena as well as, to estimate the effect of tumor growth in eloquent areas.
Main results of GLIOHAB project:
1. ONCOhabitats results for The Cancer Genome Atlas Glioblastoma Multiforme (TCGA-GBM): Segmentation and Hemodynamic Tissue Signature. (2021) doi:10.5281/zenodo.4704090.
2. ONCOhabitats results for Ivy Glioblastoma Atlas Project (Ivy Gap): Segmentation and Hemodynamic Tissue Signature. (2021) doi:10.5281/zenodo.4704106.
3. Efficiency of deep learning on segmenting longitudinal postoperative glioblastomas. ECR 2020 EPOS
2020(si apre in una nuova finestra).
4. CBV BRAIN ATLAS v1. (Zenodo, 2021). doi:10.5281/zenodo.4757126.
5. JOncohabitats glioma segmentation model. Lect. Notes Comput. Sci. Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinforma. 11992 LNCS, 295–303 (2020).
6. The use of a DSC-MRI perfusion atlas for cerebral blood volume normalization and its impact in improving prognostic estimation (Ref. 1087). in (2021).
7. The impact of EPI-based distortion correction of dynamic susceptibility contrast MRI on cerebral blood volume estimation in patients with glioblastoma. Eur. J. Radiol. 132, 109278 (2020).
8. Non-local spatially varying finite mixture models for image segmentation. Stat. Comput. 31, (2021).
9. Robust association between vascular habitats and patient prognosis in glioblastoma: An international multicenter study. J. Magn. Reson. Imaging 51, 1478–1486 (2020).
10. CHigher vascularity at infiltrated peripheral edema differentiates proneural glioblastoma subtype. PLOS ONE 15, e0232500 (2020).
11. Detection of local microvascular proliferation in IDH wild-type Glioblastoma using relative Cerebral Blood Volume. medRxiv 2021.04.19.21255589 (2021) doi:10.1101/2021.04.19.21255589.
12. Differential effect of vascularity between long- and short-term survivors with IDH1/2 wild-type glioblastoma. NMR Biomed. (2021)
13. MGMT methylation may benefit overall survival in patients with moderately vascularized glioblastomas. Eur. Radiol. (2020)
