Subscribe to eTOC

Voxel-based MRI Analysis Detects Unseen Lesions in Focal Epilepsy

A novel voxel-based MRI method that assessed multiple modalities helped identify lesions in patients with focal epilepsy that had been missed on traditional MRI, according to a study presented at AES2020, the virtual annual meeting of the American Epilepsy Society.

The technique identified lesions in approximately two-thirds of those who previously had had negative MRI scans.

“The hope would be to have this implemented far down the road in helping to detect lesions in MRI negative patients," said Jonah Isen, BComp, an undergraduate researcher at Queen's University in Ontario, Canada, who presented the findings. “At this point, if you're MRI-negative and medicine is not working to help curb your epilepsy, then you may need some sort of invasive testing to figure out where the lesion is so that you can proceed with surgery. So this will hopefully eliminate the step of invasive testing."

Voxel-based morphometry—which uses three-dimensional data points to assess the brain—and the MRI modalities used are not new. But the investigation of  several modalities in a comparative study is novel, he continued.

“Some of these modalities hadn't been researched before for voxel-based analysis," Isen said.

Subjects included 62 healthy controls, 44 people with epilepsy and lesions visible on MRI scans, and eight people with epilepsy whose MRI scans appeared normal.

About a third of the time, the MRI scans of people with refractory focal epilepsy appear normal, Isen explained.

The experimental protocols involved a basic T1-MRI scan, along with diffusion tensor imaging (DTI) and neurite orientation dispersion and density imaging (NODDI), an advancement of diffusion weighted imaging that allows better visuals of microstructures such as dendrites and axons.

The different images are co-registered to the person's T1 scan, the most detailed of the scans. Then, through a process of normalization, the images' sizes and shapes are morphed to a template to make an apples-to-apples comparison possible.

“Once the brains are the same shape and the same size, it's a matter of performing statistical tests on them to see which areas of the brain are statistically different from the control brains," Isen said.

Researchers found that the decreased neurite density index (NDI)—the image that reflected a decrease in the signal intensity of NDI compared to a regular brain—produced the most useful imaging. NDI, a parameter derived from NODDI, refers to imaging that represents the density of neurites. This modality accurately detected 76 percent of the lesions of the epilepsy patients with visible lesions on MRI and 63 percent of the lesions of those with epilepsy who had had an MRI scan that appeared normal.

Their next step is to detect lesions using the combination of several modalities at once, in the hope that the accuracy of detection improves even further, Isen said. The approach is far from being able to be used clinically, but these findings represent an important step in that direction, he said.

Kathryn Davis, MD, assistant professor of neurology at the University of Pennsylvania Perelman School of Medicine, who was not involved with the study, said the approach is promising.

“If NODDI is able to identify a clear lesion on MRI, a more directly therapeutic approach could be taken," she said. “For instance, a patient may be able to have more targeted intracranial EEG placement or proceed directly to surgical intervention."

But she said the technique needs more study.

“It is feasible to obtain NODDI imaging on clinical 3T MRI scanners. However, the scan times are longer than typical clinical scans," she said. “The data presented here is from a very small number of patients and it is unclear what the false positive rate is for lesion detection with NODDI in patients without visible lesions on conventional MRI. A multicenter study including a larger number of patients is needed to determine the scalability of NODDI in clinical care."

Orrin Devinsky, MD, FAAN, the director of the New York University Langone Comprehensive Epilepsy Center, agreed that the findings were a positive step.

These findings “are encouraging and reflect a growing interest in advancing imaging and machine-learning techniques to improve detection of subtle abnormalities that escape detection by the human eye and mind," he said.

“This study highlights progress using voxel-based morphology and the neural density index.  The key for this and similar work is validating this in patients who undergo epilepsy surgery to resect or ablate or stimulate the areas identified and compare long-term outcomes to those in which no such lesions are found," Dr. Devinsky told Neurology Today At the Meetings.

Isen had no relevant disclosures.

Link Up for More Information

AES Abstract 67: Isen J, Perera-Ortega A, Vos S, et al. Voxel based analysis of multimodal MRI for brain lesion detection in focal epilepsy.