Brain-computer interfaces (BCIs) are an emerging research area that might profoundly benefit people with severe motor loss and have potential applications in other situations where physical and voice interfaces are impractical. This project will develop techniques for using ultrasound imaging to measure and interpret brain activity and integrate ultrasound derived information with existing electroencephalography (EEG)-based techniques. Doing this will provide faster and more precise measures of brain activity compared to existing techniques. It will also allow for the integration of sub-cortical information into BCIs, information that EEG systems cannot provide. To achieve this goal, portable ultrasound helmet prototypes will be developed and solutions to transcranial ultrasound image quality problems will be resolved, which will also prove useful for medical ultrasound more generally. The helmet will advance the feasibility of mobile brain activity measurements. Along with the development and prototyping of an ultrasound BCI device, the development of new algorithms to integrate ultrasound and EEG signals will lead to new techniques for both understanding brain activity and using it to interact with computers. The work will also provide the basis for modernizing and creating several new courses around the intersection of machine learning and biomedical signal processing, as well as providing research opportunities for outreach programs that involve high school students from underrepresented groups in STEM.
The project will be organized around several main activities. The first activity is to develop high-dynamic range ultrasound imaging methods for transcranial ultrasound. This will include the integration of multiple ultrasound transducer arrays and development of a new beamforming strategy utilizing iterative multi-stage regularization supported by deep learning methods to remove high-amplitude noise while preserving small signal variations corresponding to blood flow. The second activity is to use the imaging methods to support an adaptive demodulation technique enabling non-contrast functional ultrasound analyses without contrast agents at low transducer frequencies appropriate for human transcranial imaging. The third activity is to develop a wearable helmet that captures both ultrasound and EEG signals together. The ultrasound and EEG will be physically registered using a new probe that will serve as both an EEG and an ultrasonic point source. The helmet data will be used to develop both ultrasound-only and integrated (EEG plus ultrasound) algorithms for detecting brain states that computer systems can use either as direct commands or as input to adapt their behavior to a user's current brain state.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
