NIDCD/ORD Workshop on Brain Computer Interfaces for Speech Synthesis
The National Institute on Deafness and Other Communication Disorders and the NIH Office of Rare Diseases jointly sponsored a workshop to evaluate the potential for brain-computer interfaces (BCI) to provide a means for speech synthesis and control of other forms of assistive technology that support communication in patients who are locked in. The workshop was held May 26-27 at the Bethesda Hyatt Regency Hotel. Sixteen scientists and clinicians attended to discuss topics ranging from brain computer interface data acquisition and analysis, cortex neurophysiology, speech synthesis, augmentative and alternative communication, and patient care needs for the locked-in patient.
The meeting began with an overview of the practices currently used to provide augmentative and alternative communication for patients. These techniques make effective use of existing technology, such as laser pointers, and residual movement control in patients. Other presentations highlighted the potential for advancement in this area by pioneering patient studies. For example, one patient has already been implanted with a recording electrode in Broca’s area of the cortex. Research and development activities are underway to create a means for this patient to control speech synthesis software, based on signals recorded from clusters of neurons. Other groups have developed an approach based on an array of electrodes that has been implanted in a patient’s cortex. The activity of cortical neurons is varied by the patient and control over the horizontal and vertical movement of a computer cursor has been demonstrated. Further research and development in these areas holds great promise for the development of a high bandwidth controller that could be used to benefit both paralyzed and locked-in individuals.
Useful control signals can also be obtained from electrodes placed near the surface of the cortex or outside of the skull; these techniques were reviewed at the workshop and their relative merits discussed. Signals are based on the correlated activity taken from many cortical neurons in this approach. Trained users using electroencephalograph electrodes have demonstrated the ability to control the vertical and horizontal movement of a computer curser to achieve a variety of tasks, including the control of spelling software.
Although the field is poised to make substantial progress based on the use of BCIs as they reach the stage of translation from the laboratory benchtop, substantial obstacles remain. The goal of creating software capable of extracting the intended action and/or communication, based on the neural response acquired, is likely to be a challenging problem. A number of different algorithms that could be used to decode the patient’s intended speech, based on changes in neural activity, were discussed by experts in this area of statistics and engineering. Algorithm development for control signals will be a substantial challenge in this area, but a number of approaches are already in development and are likely to accelerate efforts.
The basic acoustic parameters which vary during normal speech production were also reviewed by experts in this field. Issues that affect the intelligibility of speech in patients with damage to the larynx and laryngeal nerve were discussed. Intelligent spelling algorithms were identified as useful tools in this area, such as the Dasher software which assists the users by taking into account the relative probability of subsequent letters and words based on those preceding them. Development efforts would also need access to shared software libraries that provide communication capabilities that scale up in capability according to the quality of the control signals. The need for shared resources extends to the raw data acquired from cortex neurons in subjects attempting to speak or speaking normally; various means of sharing available data across groups interested in this new area of research were explored.
In the concluding discussions, there was widespread agreement that research and development in this area would benefit from input across a number of interdisciplinary groups. NIDCD is actively encouraging continued development in this area through the NIDCD Translational Research funding opportunity announcements and investigator initiated R01 grant applications. Support for continued meetings that foster development of this technology as a communication device is also likely to be useful; the meeting participants were encouraged to formulate plans for further conferences specifically devoted to this topic.