Synthetic MRI May Help Reduce Gadolinium Use in Brain Tumor Management

By Brandon May

November 08, 2021

NEW YORK (Reuters Health) - Researchers in Germany have developed a deep-learning-based tool to generate synthetic post-contrast T1-weighted magnetic resonance imaging (MRI) that allows assessment of treatment response in patients with brain tumors.

The findings may help reduce the need for gadolinium-based contrast agents, researchers say in a paper in The Lancet Digital Health.

"Gadolinium is really important for glioma grading and monitoring, but repeated use comes with increased costs, patient burden, and potential safety concerns," Dr. Frederik Barkhof of Amsterdam University Medical Center, who co-authored an accompanying editorial, told Reuters Health by email.

Dr. Philipp Vollmuth of the Heidelberg University Hospital in Germany and colleagues evaluated the diagnostic value and feasibility of synthetic post-contrast T1-weighted MRI produced from pre-contrast MRI sequences for assessment of tumor response.

They trained and validated a deep convolutional neural network (dCNN) for synthesizing these sequences based on MRI data from 1,486 patients being treated for glioblastoma. For independent testing, they used MRI data from 521 MR.

The team reports that "the use of synthetic post-contrast T1-weighted sequences for volumetric quantification of the contrast-enhancing tumour burden and assessment of the treatment response in independent large-scale test sets yielded no significant difference in the time to progression as compared with using true post-contrast T1-weighted sequences with administration of gadolinium-based contrast agents."

"Moreover," they add, "the strength of association between the time to progression as surrogate endpoint for predicting the patients' overall survival was similar when derived from synthetic versus true post-contrast T1-weighted sequences."

Dr. Sanjay Aneja, an assistant professor in the department of therapeutic radiology at Yale School of Medicine, in New Haven, Connecticut, told Reuters Health by email, "Patients with brain tumors often are subject to repeated scans every 6 to 8 weeks and although the side effects of contrast are rare, the cumulative effects can be of concern."

"Any intervention that can provide effective imaging of the brain without subjecting patients to contrast would be very valuable," said Dr. Aneja, who wasn't involved in the new research.

"While the study has shown the effectiveness of AI algorithms to reduce contrast exposure for primary brain tumors which are often large and easier to identify, the utility of this method on smaller brain tumors or cancer that has spread to the brain is still unknown," he added. "Moreover, whether physicians would feel comfortable making treatment decisions based on MRI images which have been altered by AI algorithms is still unknown."

Dr. Barkhof, the editorialist, said that further study is needed to determine how well these types of scans perform in differentiating low-grade and high-grade tumors. "Additionally, clinical implementation and adoption has not been tested," he said.

Dr. Vollmuth did not respond to an interview request.

SOURCE: The Lancet Digit Health, online October 20, 2021.