An magnetic resonance imaging (MRI) radiomics prediction model can stably predict the survival of patients with glioma, according to a study published in Brain. The model could potentially aid with clinical prognosis prediction and future decision-making for immunotherapy.

Highly lethal, glioma is the most common primary cancer in the central nervous system. An accurate prognosis guides postoperative treatment. While MRI is widely used for identifying size and location of glioma, MRI could be utilized as a prognostic or predictive biomarker if a predictive model was developed for clinical practice.

The objective of the current study was to develop and validate an MRI radiomics prediction model through a machine learning-based method to predict the survival of patients with glioma.


Continue Reading

Researchers conducted a retrospective, prospective cohort study of 652 patients with glioma: the discovery cohort of 167 patients with gliomas collected retrospectively from Beijing Tiantan Hospital; the external validation cohort of 261 patients with gliomas from The Cancer Genomae Atlas database, accessed December 2021; the prospective validation cohort of 224 patients with gliomas from the prospective study from Beijing Tiantan Hospital, enrolled from November 2016 to August 2019.

Patients in the validation cohort were followed-up trimonthly by clinic or telephone for an average of 709 days. Ribonucleic acid (RNA)-sequencing data, single-cell sequencing, and immunohistochemical staining of glioma from the discovery and external validation cohorts established the relationship between biological function and key radiomics features.

Applying a standard multivariate approach, the prediction model of 14 radiomic features, constructed from preoperative T2-weighted MRI images of 1731 radiomic features in the discovery cohort, displayed a high predictive power for overall survival in the external and prospective validation cohorts of patients with glioma.

The prediction model radiomic features were associated with immune response, and tumor macrophage infiltration.

This study may be limited by the significantly distinct difference between cohorts in patient age, isocitrate dehydrogenase 1 (IDH1)gene, and chromosome 1p/19q pathognomonic biomarker. According to researchers, “the prediction efficiency might not be significantly affected by the difference between the discovery and validation cohorts.”

The researchers stated, “A radiomics prediction model that incorporated the clinical prognosis prediction and tumor immune microenvironment assessment was established to change the current clinical management of patients with gliomas.”

This prediction model is an independent factor from traditional prognostic factors for gliomas but can be incorporated with those factors to predict prognosis for patients with gliomas.

“The radiomics prediction model has the advantages of non-invasive, economical, and can guide the clinical treatment of glioma before surgery,” the researchers noted.

Reference

Li G, Li L, Li Y, et al. An MRI radiomics approach to predict survival and tumor-infiltrating macrophages in gliomas. Brain. Published online February 6, 2022. doi:10.1093/brain/awab340