We aimed to construct and validate a radiomics prediction model based on preoperative T2-weighted MRI for prognosis and chemosensitivity prediction in patients with glioma.

A total of 576 glioma patients were enrolled in this study. The training and validation group included 324 patients and 127 patients respectively with preoperative MRI image data, tumor transcriptome sequencing data and clinical information. The prospective validation group consisted of 125 patients with preoperative MRI image data and clinical information. The radiomics prediction model was constructed based on the prognostic relevant radiomic features of glioma patients in the training group. The radiomics prediction model was validated inpatients of retrospective and prospective validation groups. Functional annotation of radiomic features was performed by pearson correlation analysis of biological process scores and radiomic features values of patients in the training group and validated by transcriptome sequencing, single cell sequencing, reactive oxygen species detection and endoplasmic reticulum stress detection of tumors of patients in retrospective and prospective validation groups.

The radiomics prediction model, which consisted of 17 radiomic features, showed highly predictive stability in overall survival and progression-free survival prediction. Compared with patients who underwent postoperative radiotherapy alone, only patients in the high-risk group benefited from postoperative chemoradiotherapy. The prognostic relevant radiomic features were closely related to immune response, reactive oxygen species metabolism and endoplasmic reticulum stress of glioma cells in silico and in vitro.

The radiomics prediction model serves as a non-invasive tool to predict prognosis and temozolomide chemosensitivity of glioma patients based on preoperative T2-weighted MRI.