Deep-Learning Convolutional Neural Networks Accurately Classify Genetic Mutations in Gliomas

Area of Science:

• Neuro-oncology
• Medical Imaging
• Artificial Intelligence

Background:

• The World Health Organization emphasizes integrating genetic information for glioma classification.
• Tissue sampling is standard, but noninvasive imaging may offer complementary genetic insights.
• Glioma subtypes have distinct genetic profiles influencing treatment and prognosis.

Purpose of the Study:

• To train a convolutional neural network (CNN) to predict molecular genetic mutation status in gliomas noninvasively.
• To identify key MR imaging features predictive of specific genetic mutations.
• To assess the accuracy of AI in classifying glioma genetic subtypes.

Main Methods:

• Retrospective analysis of 259 glioma patient MR imaging and molecular data from The Cancer Imaging Archives.
• Training a CNN to classify isocitrate dehydrogenase 1 (IDH1) mutation, 1p/19q codeletion, and O6-methylguanine-DNA methyltransferase (MGMT) promotor methylation.
• Utilizing principal component analysis on the CNN's final layer to extract critical imaging features.

Main Results:

• High classification accuracy achieved: IDH1 mutation (94%), 1p/19q codeletion (92%), and MGMT promotor methylation (83%).
• Distinct MR imaging features correlated with each genetic category, including tumor margin definition, edema, necrosis, and textural patterns.
• CNN successfully learned key imaging components without prior feature selection.

Conclusions:

• Machine learning, specifically CNNs, can accurately classify genetic mutations in low- and high-grade gliomas using MR imaging data.
• AI models can identify predictive imaging biomarkers for glioma genetics, enhancing noninvasive diagnostic capabilities.
• This approach demonstrates the potential of AI in advancing precision neuro-oncology by integrating imaging and molecular data.