Search

Background To develop a radiomics signature for predicting overall survival (OS)/progression-free survival (PFS) in patients with medulloblastoma (MB), and to investigate the incremental prognostic value and biological pathways of the radiomics patterns. Methods A radiomics signature was constructed based on magnetic resonance imaging (MRI) from a training cohort (n = 83), and evaluated on a testing cohort (n = 83). Key pathways associated with the signature were identified by RNA-seq (GSE151519). Prognostic value of pathway genes was assessed in a public GSE85218 cohort. Findings The radiomics-clinicomolecular signature predicted OS (C-index 0.762) and PFS (C-index 0.697) better than either the radiomics signature (C-index: OS: 0.649; PFS: 0.593) or the clinicomolecular signature (C-index: OS: 0.725; PFS: 0.691) alone, with a better calibration and classification accuracy (net reclassification improvement: OS: 0.298, P = 0.022; PFS: 0.252, P = 0.026). Nine pathways were significantly correlated with the radiomics signature. Average expression value of pathway genes achieved significant risk stratification in GSE85218 cohort (log-rank P = 0.016). Interpretation This study demonstrated radiomics signature, which associated with dysregulated pathways, was an independent parameter conferring incremental value over clinicomolecular factors in survival predictions for MB patients. Funding A full list of funding bodies that contributed to this study can be found in the Acknowledgements section.

The 2016 WHO classification of central nervous system tumors has included four molecular subgroups under medulloblastoma (MB) as sonic hedgehog (SHH), wingless (WNT), Grade 3, and Group 4. We aimed to develop machine learning models for predicting MB molecular subgroups based on multi-parameter magnetic resonance imaging (MRI) radiomics, tumor locations, and clinical factors. A total of 122 MB patients were enrolled retrospectively. After selecting robust, non-redundant, and relevant features from 5,529 extracted radiomics features, a random forest model was constructed based on a training cohort (n= 92) and evaluated on a testing cohort (n= 30). By combining radiographic features and clinical parameters, two combined prediction models were also built. The subgroup can be classified using an 11-feature radiomics model with a high area under the curve (AUC) of 0.8264 for WNT and modest AUCs of 0.6683, 0.6004, and 0.6979 for SHH, Group 3, and Group 4 in the testing cohort, respectively. Incorporating location and hydrocephalus into the radiomics model resulted in improved AUCs of 0.8403 and 0.8317 for WNT and SHH, respectively. After adding gender and age, the AUCs for WNT and SHH were further improved to 0.9097 and 0.8654, while the accuracies were 70 and 86.67% for Group 3 and Group 4, respectively. Prediction performance was excellent for WNT and SHH, while that for Group 3 and Group 4 needs further improvements. Machine learning algorithms offer potentials to non-invasively predict the molecular subgroups of MB.

Background: The 2016 WHO Classification of Central Nervous System tumors has included four molecular subgroups under medulloblastoma (MB) as sonic hedgehog (SHH), wingless (WNT), Grade 3 and Group 4. We aimed to develop machine learning models for predicting MB molecular subgroups based on multimodal magnetic resonance imaging (MRI) radiomics, tumor locations and clinical factors. Methods: 122 MB patients were enrolled retrospectively. After selecting robust, nonredundant and relevant features from 5529 extracted radiomics features, a random forest model was constructed based on a training cohort (n = 92) and evaluated on a testing cohort (n = 30). By combining the radiographic features and clinical parameters, two combined prediction models were also built. Findings: The subgroup can be classified using an 11-feature radiomics model with a high area under curve (AUC) of 0.8264 for WNT and modest AUCs of 0.6683, 0.6004 and 0.6979 for SHH, Group 3 and Group 4 in testing cohort, respectively. Incorporating location and hydrocephalus into the radiomics model resulted in improved AUCs of 0.8403 and 0.8317 for WNT and SHH, respectively. After adding gender and age, the AUCs for WNT and SHH were further improved to 0.9097 and 0.8654, while the accuracies were 70% and 86.67% for Group 3 and Group 4, respectively. Interpretation: The prediction performance was excellent for WNT and SHH, while that for Group 3 and Group 4 needs further improvements. Machine learning algorithms offer potentials to non-invasively predict molecular subgroups of MB. Funding Statement: This research was supported by the National Natural Science Foundation of China (No. 81702465, 61571432 and U1804172), the Science and Technology Program of Henan Province (No. 182102310113, 192102310123 and 192102310050). Declaration of Interests: The authors declare no potential conflicts of interest. Ethics Approval Statement: The Human Scientific Ethics Committee of the First Affiliated Hospital of Zhengzhou University has approved the protocol of this study (No. 2019-KY-176).