Artificial Intelligence

AI Deep Learning Classifies Brain Tumors from a Brain Scan

Detecting the presence of tumors and typology with high accuracy.

Posted August 24, 2021 | Reviewed by Kaja Perina

Key points

An AI deep learning model developed by the Washington University School of Medicine is able to classify brain tumors using a single 3D MRI Scan.
The new convolutional neural network (CNN) for classifying brain tumors has an accuracy rate of over 93 percent from a non-invasive MRI scan.
The AI machine learning model can be extended to include more brain tumor types and neurological disorders.

Source: ParallelVision/Pixabay

Researchers at the Washington University School of Medicine use artificial intelligence (AI) deep learning to classify common brain tumors with a high degree of accuracy using a single magnetic resonance imaging (MRI) scan. The new peer-reviewed study has been accepted for publication in Radiology: Artificial Intelligence.

“To the best of our knowledge, this is the first study to address the most common intracranial tumor-types and directly determine the tumor class as well as detect the absence of tumor from a 3D MR volume,” wrote the researchers.

Last year there were over 308,000 new cases of brain and nervous systems cancer, and more than 250,000 deaths worldwide according to the Global Cancer Statistics (GLOBOCAN) 2020 report. In the United Kingdom, over 11,000 people are diagnosed with a primary brain tumor annually, according to the National Health Service (NHS). In the United States, there are nearly 700,000 Americans who are living with a primary brain tumor, and the average survival rate for all malignant brain tumor patients is only 35 percent according to the National Brain Tumor Society 2019 report.

“MRI may be used as a compliment, or in some cases alternative, to histopathologic examination due to its noninvasive nature and high soft-tissue contrast,” the researchers wrote.

In this study, the researchers used retrospective deidentified data from the Washington University School of Medicine, in addition to the Brain Tumor Image Segmentation, The Cancer Genome Atlas Glioblastoma Multiforme, and The Cancer Genome Atlas Low Grade Glioma datasets.

To develop and test the model, the team focused on seven classifications that include a healthy class, as well as six common brain tumor types: high-grade gliomas, low-grade gliomas, brain metastases, meningiomas, pituitary adenomas, and acoustic neuromas.

There are over 150 documented brain tumors according to the American Association of Neurological Surgeons (AANS). Benign tumors include chordomas, craniopharyngiomas, gangliocytomas, glomus jugulare, meningiomas, pineocytomas, pituitary adenomas, and schwannomas.

A majority of malignant brain tumors (78 percent) are gliomas, which are named after the glia (also called neuroglia). Glial cells are non-neuronal cells that do not generate electrical impulses and are found in the central nervous system and peripheral nervous system. Nearly half of all gliomas are astrocytomas. Other gliomas include glioblastoma multiforme (GBM), ependymomas, medulloblastomas, and oligodendrogliomas.

The scientists developed a three-dimensional (3D) convolutional neural network (CNN) architecture to classify healthy and six intracranial tumor types as well as validate its performance.

“The model achieved high accuracy on a heterogeneous dataset and showed excellent generalization capabilities on unseen testing data,” wrote the researchers. “These results suggest that deep learning is a promising approach for automated classification and evaluation of brain tumors.”

With the internal testing dataset with all seven image classifications, the AI algorithm achieved an accuracy of over 93 percent, and the sensitivities ranged from 91-100 percent. The positive predictive value ranged from 85-100 percent.

“These results suggest that deep learning is a promising approach for automated classification and evaluation of brain tumors,” the researchers reported.