In the presence of continuously growing incidence of brain disorders affecting both adults (e.g. dementia) and children (e.g. attention deficit/hyperactivity disorder (ADHD)), there is a pressing need for effective treatments. An examination of new drugs and therapies requires reliable and cost-effective biomarkers to diagnose disorders and monitor their development, especially treatment related changes. The diagnostic objectivity, lowering technological costs and non-invasiveness of neuroimaging techniques such as electroencephalogram (EEG) or magnetoencephalogram (MEG) renders EEG/MEG-based biomarkers an attractive and popular choice in clinical practice. Mentis Cura, a medical research company, has undertaken a long-lasting effort of collecting EEG data recorded from patients with neurological disorders and searching for reliable electrophysiological correlates of the brain malfunction. The company has been particularly successful in diagnosis of different types of dementia and developmental disorders. In this project it is intended to develop an alternative brain-inspired computational approach to brain signal analysis and diagnostic inference. It is envisaged that the proposed method, based on an abstract model of processing and coding neural information in the brain’s neocortex, will allow for a more exploratory search for relevant spatiotemporal signal correlates, and a more flexible inference mechanism that can benefit from both supervised and unsupervised forms of learning from data. In consequence, the method is expected to robustly handle more complex EEG/MEG recording paradigms, account for different data types and sources, and produce more informative output, e.g. by providing more specific characterization of patients’ profiles rather than just differential diagnosis. The additional computational cost will be mitigated by a massively parallel implementation and simulations on a supercomputer. The clinical relevance of the developed approach to diagnosis of brain disorders will be assessed comparatively on the existing data sets using the current Mentis Cura’s tools and other state-of-the-art brain signal classification/clustering methods. Finally, we envision further implications of the project in the context of generic data analysis and computation technology, as well as diagnostics and treatment of neurological disorders.
Principle investigator: Anders Lansner
Partner: Mentis Cura (Ivar Meyvantsson)