Neural network AI identifies brain tumor types pre-surgery with high accuracy

Editor's Note

Researchers have found a way to use artificial intelligence (AI) to identify tumor types related to brain tumors prior to surgery. The findings were published on October 11 by the journal Nature in a study titled, "Ultra-fast deep-learned CNS tumor classification during surgery."

Following are some highlights, according to the study:

• Central nervous system tumors are one of the most lethal cancer types, particularly among children. 
• Primary treatment includes surgery, but surgeons have limited knowledge of the precise tumor type prior to the surgical procedure. 
• In the study, researchers developed Sturgeon, a transfer-learned neural network, to enable molecular subclassification of central nervous system tumors based on existing methylation profiles. 
• Sturgeon delivered an accurate diagnosis 40 minutes after starting sequencing in 45 out of 50 retrospectively sequenced samples (abstaining from diagnosis of the other 5 samples). 

The researchers were able to demonstrate that this AI method could be applied to real world surgical procedures, achieving a diagnostic turnaround time of less than 90 minutes in 25 surgeries, in which 18 (72%) diagnoses were correct and seven did not reach the required confidence threshold. The findings led them to conclude that the tool can be valuable in assisting neurosurgical decision-making, potentially preventing neurological comorbidity and avoiding additional surgeries.