MR Based Prediction of Molecular Pathology in Glioma Using Artificial Intelligence

Non-invasive and precise prediction for molecular biomarkers such as 1p/19q co-deletion, MGMT methylation, IDH and TERTp mutations is challenging. With the development of artificial intelligence, much more potential lies in the preoperative conventional/advanced MR imaging (T1 weighted imaging, T2 weighted imaging, FLAIR, contrast-enhanced T1 weighted imaging, diffusion-weighted imaging, and perfusion imaging) could be excavated to aid prediction of molecular pathology of gliomas. The creation of a registry for primary glioma with detailed molecular pathology, radiological data and with sufficient sample size for deep learning (>1000) provide considerable opportunities for personalized prediction of molecular pathology with non-invasiveness and precision.

Inclusion criteria

• Patients must have radiologically and histologically confirmed diagnosis of primary glioma
• Life expectancy of greater than 3 months
• Must receive tumor resection
• Signed informed consent

Exclusion criteria

• No gliomas
• No sufficient amount of tumor tissues for detection of molecular pathology
• Patients who have any type of bioimplant activated by mechanical, electronic, or magnetic devices
• Patients who are pregnant or breast feeding
• Patients who are suffered from severe systematic malfunctions