If a person does not speak but is thinking of an object, for example an apple, is it possible to decode that person's inner voice from brain signals and recognize the word apple? This is a challenging question that the research group in Machine Learning is trying to answer.
– If we succeed, this research will open doors for endless possibilities, says Foteini Liwicki, Associate Professor of Machine Learning.
– If we can help patients who suffer from neurological disorders to communicate with the real world, it will be a worthy contribution.
From data to patterns
Brain analysis using machine learning is an area of research in the machine learning and neuroscience community. By analyzing brain data, the researchers aim at detecting inner speech. For example, this could help people with Locked-In-Syndrome (LIS), that is, patients that are cognitively aware but cannot move or communicate verbally.
There are many ways of collecting brain data, for example electroencephalography (EEG), fMRI (functional magnetic resonance imaging), and near-infrared spectroscopy (fnirs). Different modalities have different types of information and most of the signals are noisy. The researchers at Luleå university of Technology use machine learning algorithms to filter the signals and to extract meaningful patterns from the data. Every result contributes with a piece to the great puzzle of how the human brain functions.
– The human brain is still unexplored territory and every day we learn new things and it is fascinating. The root cause of a lot of neurological disorders like Alzheimer's and Parkinson's, is not fully known. Therefore, any contribution of machine learning algorithms decoding this, will help humanity and society, says Foteini Liwicki.
The research group in Machine Learning is expanding within the realm of brain analysis. They recently conducted a pilot study together with researchers from Umeå University and University of Bath, using the equipment of Stockholm University Brain Imaging Centre, where they collected brain data and applied machine learning methods to analyze them. Last spring, they also participated in the Virtual Br41n.IO Brain-Computer Interface Designer’s hackathon and won the data analysis track.
– We also have national and international collaborators in data acquisition and filtering. We hope to be able to continue down this research road by creating a consortium and find more partners nationally and internationally, Foteini Liwicki concludes.