Periodic Reporting for period 1 - OsciLang (OsciLang: A neurofeedback system based on oscillatory activity for diagnosis and intervention in language and reading impairments)
Reporting period: 2018-10-01 to 2020-03-31
The goal of OsciLang was to provide an affordable, lightweight, wearable brain-computer-interface neurofeedback system to facilitate the detection and treatment of language disorders such as dyslexia and specific language impairment. To date there are no established neurofeedback protocols for treating language or reading impairments. In essence, this tool would (a) diagnose/measure and (b) improve/rehabilitate an individual’s ability to synchronize their brain’s activity with changes in a speech signal, named Brain-Speech-Entrainment (BSE). This novel type of neurofeedback, based on phase coherence of the neural oscillatory activity with speech, would allow us to detect and enhance language function. This idea capitalized on our recent research on neural oscillations during speech processing, where we found reduced auditory tracking in delta band in dyslexic readers compared with controls, confirming and advocating impaired auditory processing in dyslexia. Thus, the suggestion was that literacy skills in dyslexics will improve by means of neurofeedback training protocols that would enhance BSE capabilities, especially in noisy environments. In sum, OsciLang would help to sustainably enhance brain-speech synchronization strength by allowing participants to regulate their neural activity while tracking speech. This would be accomplished by presenting participants with a simple and friendly interface that would quantify basic parameters of their brain activity in real time, measured via EEG. This information could then be used by participants to facilitate top-down control on specific activation patterns.
However, available methods to measure BSE (e.g. phase-coherence) depend on long data segments. This introduces an important limitation for BCIs to offer a sense of real-time control to users. In this project, we aimed to develop and test novel/promising BSE methods with high temporal precision, including phase-based methods using the Hilbert transform and decoding strategies based on multivariate-regression (i.e. TRFs, temporal response functions). We produced data simulations of time-varying BSE to test the temporal sensitivity of our methods, and we have acquired several datasets to investigate BSE and the potential of BCI. However, we did not find a significant entrainment difference between the subjects who received BSE feedback from those who received beta power feedback. The present lack of evidence supporting BSE-based BCI may be due to limitations in temporal precision of the BSE metric. Overall, we acknowledge that the development of a BCI based on a complex metric such as BSE is challenging to optimize.
Finally, the methods we developed and our combined strategy to a) simulate data time-series with controlled temporal variation of neural coupling, b) testing of real EEG data, both using traditional gel-EEG and c) modern (light weight) dry-EEG systems is the appropriate direction to achieve the goals of OsciLang. We have made important progress in each of these levels.
However, available methods to measure BSE (e.g. phase-coherence) depend on long data segments. This introduces an important limitation for BCIs to offer a sense of real-time control to users. In this project, we aimed to develop and test novel/promising BSE methods with high temporal precision, including phase-based methods using the Hilbert transform and decoding strategies based on multivariate-regression (i.e. TRFs, temporal response functions). We produced data simulations of time-varying BSE to test the temporal sensitivity of our methods, and we have acquired several datasets to investigate BSE and the potential of BCI. However, we did not find a significant entrainment difference between the subjects who received BSE feedback from those who received beta power feedback. The present lack of evidence supporting BSE-based BCI may be due to limitations in temporal precision of the BSE metric. Overall, we acknowledge that the development of a BCI based on a complex metric such as BSE is challenging to optimize.
Finally, the methods we developed and our combined strategy to a) simulate data time-series with controlled temporal variation of neural coupling, b) testing of real EEG data, both using traditional gel-EEG and c) modern (light weight) dry-EEG systems is the appropriate direction to achieve the goals of OsciLang. We have made important progress in each of these levels.