Active Projects | Overview
We work extensively with the BCH Epilepsy Center on multiple projects. All of our work with the Epilepsy Center has a clear clinical focus, mostly around discovery of biomarkers for epilepsy, comorbid neurological or cognitive effects, and seizure detection. Our primary focus is on developing and implementing new methods for nonlinear EEG signal analysis, data integration, data structures for complex feature extraction, and machine learning for classification of epilepsy or seizures. Emerging work will involve large scale EEG data extraction from the BCH archives, and annotation for research. This work involves close collaboration with Dr. Tobias Loddenkemper in the BCH Epilepsy Center.
Early detection of autism, anxiety, and monitoring cognitive development
Autism spectrum disorder (ASD) is a complex and heterogeneous disorder, generally diagnosed on the basis of behavioral symptoms during the second year of life or later. We have demonstrated that biomarkers for emerging autism can be computed from EEG signals as early as 3 months of age. This work will continue with further testing in a general population of children in routine pediatric checkups. Related to this, we are working with Dr. Charles Nelson and his team to develop EEG-based analyses for measuring cognitive development. The goal of this research is to monitor children in low resource areas of the world for typical cognitive development. Finally, another project with the child psychiatrists at BCH (Drs. Bosquet-Enlow and Anthony) we are using nonlinear signal analysis to search for EEG biomarkers of chronic anxiety in children.
EEG annotation and crowdsourcing
The use of new signal processing and machine learning algorithms to find digital biomarkers in EEG signals requires annotations for EEG signals at scale and in an on-going manner. Because this is a highly time-consuming process for human experts, and there is no gold-standard by which to judge EEG annotations for epileptiform activity or other patterns, we are developing a crowdsourcing platform for EEG annotation and accuracy estimation. In related research that involves software and data engineering, we are beginning work on an EEG Biobank that will make BCH clinical EEG data available for researchers.
Biomedical informatics, signal processing, machine learning
All of our research in neuroinformatics is based on a nonlinear systems approach to analyzing EEG signals, and integration of the derived signal information with other physiological or clinical data. Research involves nonlinear dynamical systems theory, recurrence plot analysis, synchronization methods, tensor formulation of complex biomarker data, and machine learning for classification.