Transcoding software for EEG and fMRI data using a convolutional neural network

This technology is software that is able to produce functional magnetic resonance imaging (fMRI) data from EEG recordings and vice versa, allowing for cost-effective collection of neuroimaging data with high spatiotemporal resolution.

Unmet Need: Low-cost neuroimaging technique with high spatiotemporal resolution

Functional MRI (fMRI) can measure hemodynamic changes in the brain and is ubiquitously used for neuroimaging in cognitive neuroscience. While the method is noninvasive with high spatial resolution, fMRI suffers from low temporal resolution and is expensive. Conversely, electroencephalography (EEG) is a neuroimaging technique that is more cost-effective with high temporal resolution but low spatial resolution. As a result, there is great commercial and clinical potential for developing EEG-fMRI methods that leverage the advantages of both imaging modalities.

The Technology: Transcoding algorithm for fMRI and EEG

This technology is a software program that is able to reconstruct fMRI images from EEG recordings and vice versa. Simultaneously acquired EEG and fMRI data were used to train a convolutional neural network to learn the relationship between the two neuroimaging modalities. The transcoding from one method to the other can be achieved without any prior knowledge of the hemodynamics and leadfield estimates. As a result, researchers and clinicians can obtain neuroimaging data with high spatiotemporal resolution in a cost-effective manner.

Applications:

• Monitoring blood flow in the brain
• Surgical evaluation
• Cognitive neuroscience
• Monitoring of epilepsy
• Diagnostics and monitoring of neurodegenerative diseases
• Monitoring pain or response to stimuli
• Psychiatry

Advantages:

• Cost-effective
• Can convert from fMRI to EEG and vice-versa
• Transcoding can be achieved without any prior knowledge of the hemodynamics or leadfield estimates
• Accurate even under different operating conditions
• High spatiotemporal resolution
• Software can learn correlations and patterns between fMRI and EEG that were previously unknown

Lead Inventor:

Paul Sajda, Ph.D.

Patent Information:

Patent Status

Related Publications:

• Liu X, Sajda P. “A convolutional neural network for transcoding simultaneously acquired EEG-fMRI data” 2019 9th International IEEE/EMBS Conference on Neural Engineering. 2019 Mar 20: 477-482.

Tech Ventures Reference:

• IR CU20368

• Licensing Contact: Dovina Qu