Glioblastoma is the most deadly type of primary brain tumor. There has been a global interest in finding metrics with prognostic or predictive value obtained from pretreatment magnetic resonance images of glioblastoma patients. This protocol aims to obtain pretreatment high-resolution imaging data from a large number of glioblastoma patients.
The protocol was approved and implemented at Hospital General Universitario de Ciudad Real, Hospital Universitario de Salamanca, Instituto Valenciano de Oncología, Hospital Clínico San Carlos (Madrid), Hospital Regional Universitario de Málaga, Hospital Virgen de la Salud de Toledo, Hospital Doctor Negrín (Canarias) and Hospital Marqués de Valdecilla (Santander).
Our collaborators retrospectively collected clinical data, pathology data, and pretreatment MRI imaging datasets (pretreatment T1-weighted and T2/FLAIR sequences). Tumors were semi-automatically segmented and geometrical,
textural and morphological parameters studied as imaging biomarkers. 311 patients with unifocal and 87 with multifocal tumors were included in the study. A subset of the TCIA datasets of 110 patients with similar imaging quality
standards was included for the validation groups. This made a full dataset of 508 patients.
The most remarkable results are the development of mathematically-inspired prognostic and predictive metrics that outperformed state-of-the-art machine learning approaches. Also, the identification, based on mathematical models, of the contrast enhancing rim width and tumor surface irregularity as individual and combined prognosis and response biomarkers.
Results were published in 2016-2019. The GLIOMAT protocol has led to another study, the TOG protocol, which collects a much larger amount of data and will allow substantially improved findings.