Project Summary: The long-term goal of this research is to develop a new, non-invasive brain-computer interface (BCI) that will provide augmentative and alternative communication (AAC) capabilities to patients who have lost these capabilities due to severe motor impairments, such as completely locked-in syndrome (CLIS), amyotrophic lateral sclerosis (ALS), and severe cerebral palsy (CP). The proposed research project will work towards a long-term goal by developing a BCI based on brain imaging signals from high-density diffuse optical tomography (HDDOT). Some existing BCIs record electroencephalography (EEG) or electrocorticography (ECoG) signals from patients and then decode these signals into instructions for operating some element of the outside world, such as a cursor on a screen, a prosthetic limb, or a virtual keyboard. This functionality can enable communication. However, BCI generally has had limited success and capabilities in CLIS patients with EEG or has relied on invasive technology such as ECoG or intracortical recordings, which require surgical placement of electrodes on or beneath the brain surface. Although functional MRI (fMRI) has recently achieved great success with decoding items viewed or heard by subjects (e.g., distinguishing from among >100 viewed images), fMRI requires bulky, expensive equipment that cannot be employed for routine BCI for patients with severe motor- related communication deficits. In contrast, optical imaging approaches, such as near-infrared spectroscopy (NIRS), employ portable, wearable hardware. These optical systems are non-invasive and use non-ionizing, near-infrared light to create movies of blood oxygenation and therefore provide physiological information comparable to the fMRI signal. NIRS has recently been applied as an alternative to EEG BCI for decoding simple yes/no responses in CLIS patients. However, NIRS systems suffer from much-lower spatial resolution than fMRI, which renders NIRS unlikely to match the decoding capabilities of fMRI. High-density diffuse optical tomography (HDDOT) combines the lightweight, low-cost equipment benefits of EEG and NIRS with higher spatial resolution closer to that of fMRI at the brain surface. Recent advances in HDDOT systems have enabled average spatial localization errors <5 mm and spatial resolution <17-20 mm (substantially better than NIRS). Studies have demonstrated detailed maps of both visual and language tasks. These properties make HDDOT an ideal candidate tool for decoding brain function. The fellowship training will provide a strong foundation in optical neuroimaging methods, machine learning, and brain-computer interface. These experiences will prepare the applicant exceptionally well for a career in biomedical engineering research and for developing technology that will improve these patients’ quality of life.