Science news | Brain tumors can now be classified with a single MRI exam using the deep learning model
Washington [US]Aug 11 (ANI): In a new study, researchers developed a deep learning model capable of classifying a brain tumor into one of six common types with a single 3D MRI scan.
The study by researchers at Washington University School of Medicine was published in Radiology: Artificial Intelligence.
Also read | “Virginity Tests” on Female Recruits by Indonesian Army Stopped; Human rights groups welcome the decision.
“This is the first study looking at the most common intracranial tumors that directly determines tumor class or the absence of a tumor from a 3D MRI volume,” said Satrajit Chakrabarty, MS, PhD student led by Aristeidis Sotiras, PhD, and Daniel Marcus, PhD, in the Mallinckrodt Institute of Radiology’s Computational Imaging Lab at Washington University School of Medicine in St. Louis, Missouri.
The six most common types of intracranial tumors are high-grade gliomas, low-grade gliomas, brain metastases, meningiomas, pituitary adenomas, and acoustic neuromas. Each has been documented by histopathology, which requires the surgical removal of tissue from the site of the suspected cancer and examination under a microscope.
Also read | Atmanirbhar Narishakti Se Samvad ‘: PM Narendra Modi is scheduled to interact with women’s support groups tomorrow.
According to Chakrabarty, machine and deep learning approaches using MRI data could potentially automate the detection and classification of brain tumors.
“Non-invasive MRI can be used as an adjunct or in some cases as an alternative to histopathological examination,” he said.
To build their machine learning model, known as the convolutional neural network, Chakrabarty and researchers at the Mallinckrodt Institute of Radiology developed a large, cross-institutional data set of intracranial 3D MRI scans from four publicly available sources.
In addition to the in-house data, the team received preoperative, post-contrasting T1-weighted MRI scans from the image segmentation of brain tumors, the cancer genomatlas Glioblastoma Multiforme and the cancer genomatlas Low-Grade Glioma.
The researchers divided a total of 2,105 scans into three sub-data sets: 1,396 for training, 361 for internal tests and 348 for external tests.
The first set of MRI scans were used to train the convolutional neural network to differentiate between healthy scans and scans with tumors, and to classify tumors by type. Researchers evaluated the model’s performance using data from the internal and external MRI scans.
Using the internal test data, the model achieved an accuracy of 93.35 percent (337 out of 361) across seven imaging classes (one healthy class and six tumor classes).
The sensitivities ranged from 91 to 100 percent and the positive predictive value – or the likelihood that patients with a positive screening test will actually have the disease – ranged from 85 to 100 percent.
Negative predictive values - or the likelihood that patients with a negative screening test would actually not have the disease – ranged from 98 to 100 percent across all grades. The network attention overlapped with the tumor areas in all tumor types.
For the external test dataset, which included only two tumor types (high-grade glioma and low-grade glioma), the model had an accuracy of 91.95 percent.
“These results suggest that deep learning is a promising approach for the automated classification and assessment of brain tumors,” said Chakrabarty. “The model achieved high accuracy on a heterogeneous data set and showed excellent generalization capabilities on invisible test data.”
Chakrabarty said the 3D deep learning model moves closer to the goal of end-to-end, automated workflow by improving existing 2D approaches that require radiologists to manually delineate or characterize the tumor area on an MRI scan prior to machine processing. The Convolutional Neural Network eliminates the tedious and labor-intensive step of tumor segmentation prior to classification.
Dr. Sotiras, a co-developer of the model, said it can be extended to other types of brain tumors or neurological diseases, which may offer a way to improve much of the neuroradiological workflow.
“This network is the first step in developing an artificial intelligence-enhanced radiology workflow that can support image interpretation by providing quantitative information and statistics,” added Chakrabarty. (ANI)
(This is an unedited and auto-generated story from the Syndicated News Feed. LatestLY Staff may not have changed or edited the content.)