Predicting molecular subtype of lower-grade gliomas using machine learning

Takeaway

  • The proposed model had an accuracy of 68.7% in preoperatively diagnosing the molecular subtype of lower-grade gliomas (LGG) based on multi-modality data to predict 3-group classification.

Why this matters

  • LGG can be classified into 3 molecular subtypes depending on the presence of certain isocitrate dehydrogenase (IDH) gene mutations; each subtype is associated with different prognosis and recurrence characteristics.

  • Preoperatively identifying subtype may improve clinical outcomes, as the subtype predicts how much resection should be achieved during surgery, a factor that impacts prognosis.