Get started

Please fill in the form below and an Ezra representative will contact you within 1 business day.

Notify me

Please fill in the form below and we will be in touch when we launch in your city.

Blog / Artificial Intelligence, Brain Health, Healthcare Trends

Deep Learning Aids in Detecting a Type of Brain Aneurysm

Jun. 19 2019 by Sheherzad Raza Preisler Blog Editor, PR, & Social Media Coordinator
Deep Learning Aids in Detecting a Type of Brain Aneurysm

A study published in JAMA Network Open earlier this month investigated the effectiveness of a deep-learning algorithm called HeadXNet in helping detect intracranial aneurysms. An intracranial aneurysm, aka a cerebral aneurysm, is a weakened portion of an arterial wall in the brain that bulges abnormally. Because this spot is now compromised, it runs the risk of rupture, which is dangerous. Causes of cerebral aneurysms are not yet fully understood, making their prediction and detection incredibly important.

The team, which is based at Stanford and led by Allison Park, trained their segmentation model based on deep learning to come up with predictions for intracranial aneurysms on a type of CT scans known as CT angiography (CTA). During the testing process, eight clinicians–including six board-certified radiologists, a surgeon, and a radiology resident–saw a statistically significant rise in accuracy, sensitivity, and consensus among the various judgments made; what’s more is these advantages were brought with no compromise in diagnostic time or specificity in the CTA scan. The study showed that in the detection of intracranial aneurysms, HeadXNet improved clinician specificity by about 1.5%, sensitivity by about 6%, and accuracy by about 4%.

While CTA is the primary imaging technique used to diagnose, monitor, and plan surgeries for intracranial aneurysms, analyzing the results is incredibly time-consuming, even for the most experienced neuroradiologists. Furthermore, the results often have a low interrater agreement. In other words, they are rarely agreed upon across the board.

To the authors, these findings suggest that integrating a diagnostic model assisted by artificial intelligence (AI) could help clinicians perform better by helping them make more accurate predictions and, ultimately, take better care of their patients. Furthermore, the authors said, the AI could provide the neurosurgeons, radiologists, and other clinicians with a diagnostic tool that’s immediately applicable.

According to the Stanford team, their next steps will likely involve further study using imaging data collected from other hospitals and institutions.