Objectives: To construct a radiomics nomogram for the individualized estimation of the survival stratification in glioblastoma (GBM) patients using the multiregional information extracted from multiparametric MRI, which could facilitate the clinical decision-making for GBM patients.
Materials and methods: A total of 105 eligible GBM patients (57 in the long-term and 48 in the short-term survival groups, separated by an overall survival of 12 months) were selected from the Cancer Genome Atlas. These patients were divided into a training set (n = 70) and a validation set (n = 35). Radiomics features (n = 4000) were extracted from multiple regions of the GBM using multiparametric MRI. Then, a radiomics signature was constructed using least absolute shrinkage and selection operator regression for each patient in the training set. Combined with clinical risk factors, a radiomics nomogram was constructed based on a multivariate logistic regression model. The performance of this radiomics nomogram was assessed by calibration, discrimination, and clinical usefulness.
Results: The radiomics signature consisted of 25 selected features and performed better than clinical risk factors (i.e., age, Karnofsky performance status, and treatment strategy) in survival stratification. When the radiomics signature and clinical risk factors were combined, the radiomics nomogram exhibited promising discrimination in the training (C-index, 0.971) and validation (C-index, 0.974) sets. The favorable calibration and decision curve analysis indicated the clinical usefulness of the radiomics nomogram.
Conclusions: The presented radiomics nomogram, as a non-invasive prediction tool, could exhibit a favorable predictive accuracy and provide individualized probabilities of survival stratification for GBM patients.
Key points: • Non-invasive survival stratification of GBM patients can be obtained with a radiomics nomogram. • The proposed nomogram constructed by radiomics signature selected from 4000 radiomics features, combined with independent clinical risk factors such as age, Karnofsky performance status, and treatment strategy. • The proposed radiomics nomogram exhibited good calibration and discrimination for survival stratification of GBM patients in both training (C-index, 0.971) and validation (C-index, 0.974) sets.
Keywords: Glioblastoma; Machine learning; Multiparametric MRI; Nomogram.