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 / Brain Health, Cancer

Using Machine Learning to Classify Brain Tumors

Jan. 29 2020 by Sheherzad Raza Preisler Blog Editor
Using Machine Learning to Classify Brain Tumors

A team of researchers at Osaka University in Japan have put together a new approach to look at brain tumors. It involves marrying magnetic resonance imaging (MRI) and machine learning to quickly predict genetic mutations in glioma tumors; these types of tumors develop in the spine or brain. Their method could help patients dealing with glioma undergo a more effective treatment quicker, hopefully yielding more positive outcomes as well. Their research and results were published in Scientific Reports in late 2019.

Gliomas form in the cells that support the brain. There are two major mutations that are considered “especially important”; they are alterations in the genes for the promoter region of telomerase (TERT) and the enzyme isocitrate dehydrogenase (IDH). Pinning down these mutations can be incredibly helpful in choosing the appropriate treatment approach for patients with gliomas. The machine learning algorithm developed by the team in Osaka University may predict which mutations are found in a given case using only images of the tumors captured using MRI.

In the recent past, there has been a movement in the realm of oncology to sequence tumor cells to find the exact genetic mutations in each; this approach is inspired by the knowledge that every iteration of cancer is unique. However, some types of cancer–specifically those that develop in the brain–are difficult to assess, because a biopsy would need to be taken during surgery; this process could delay treatment in a major way.

The researchers put together the algorithm using a convolutional neural network that can isolate features from images taken during an MRI scan. They then used a type of machine learning known as support vector machines to separate patients into groups depending on the absence or presence of certain mutations. “We hope to expand this approach to other types of cancer,” remarked senior author Haruhiko Kishima, “so we can take advantage of the large cancer gene databases already collected.”

This method may be able to allow physicians to bypass surgical tissue sampling. Furthermore, it could result in more positive clinical outcomes for glioma patients, because personalized medication regimens could be administered quicker and with more ease.