New Pattern Recognition Methods in Detection and Classification of Events in the EEG of Preterm Babies
Help will soon be available for the investigation of brain activity of prematurely born babies in intensive care. That help will be provided by the new pattern recognition methods developed at the Department of Signal Processing and Acoustics of the Aalto University School of Electrical Engineering.
– It is important to identify the SAT (spontaneous activity transients) events in EEGs. These powerful, spontaneous bursts in electrical activity can take place in the brain of a prematurely born child. A lack of bursts or their limited number may in fact indicate that the brain of the premature infant is not developing well, or is not as it should be. One needs to react quickly in cases like that, explains Professor of Signal Processing and Acoustics Unto K. Laine.
New pattern recognition methods help us in spotting repetitive structures in brain EEGs, yielding information, for example, about the number of SAT events.
According to Laine, previously it has been laborious to analyze the structure of EEG signals, since those analyses have been based on visual inspection mainly. The new pattern recognition methods, on the other hand, can be used to discover even unknown repetitive structures in noisy data sequences.
Sometimes researchers hardly know what kinds of patterns can be found in a certain time series, but the methods learn to identify them automatically provided that the length of the time series is sufficient.
The method may in future move from the world of research to hospitals, because the research project will be carried out in close collaboration with the HUS Children's Hospital (in the Hospital District of Helsinki and Uusimaa) and the Pienestä kiinni ("Nearly there") project.
– The idea about co-operation has its roots in the deliberations, at the Aalto University Department of Signal Processing and Acoustics, over where and how to apply the new methods, and we also needed methods of analysis for our new EEG signals. It was this that motivated us who are involved in the research of infant's brain and made us commit ourselves to this project, explains Sampsa Vanhatalo, the physician-in-chief at the Clinical Neurophysiology Unit of the HUS Children's Hospital.
– In practice, we store the signals digitally to a computer and analyze them, but it is possible to convert the algorithm run on the computer into an application which can be integrated directly into the EEG device, clarifies Project researcher Okko Räsänen.
3D sound to help locate brain impulses
The research project is also currently testing the capabilities of 3D sound in recognizing patterns in EEG data. According to Okko Räsänen, conversion of EEG into an audible format has taken place also in the past, but no conversion into three-dimensional sound has been made.
When the EEG data is examined just as a family of curves, it is difficult to figure out the part of the brain where the impulses are. – If the data were converted into 3D sound, we could hear, for example, events in the front part of the frontal lobe, explains professor Laine.
Thus the researcher could listen to the patient's brain activity in a three-dimensional setting, just like sitting inside a giant's head. Data converted into 3D sound would become perceivable more rapidly, and one could easily return back to the important events in the brain. In addition, it would be possible to discern how brain activations move from one location to another.
The research project titled New pattern recognition methods for detecting and classifying EEG events of prematurely born babies is funded by Tekes and ends in February 2012.
Text: Tea Kalska
The steering group of the project consists of:
Mega Electronics Ltd. (Kuopio) – Arto Remes / Jukka Kinnunen
Nokia Research Center (Tampere) – Jukka P. Saarinen
University of Oulu, Department of Physics – Prof. Matt Weckström
Docent Sampsa Vanhatalo, Physician-in-chief at the Clinical Neurophysiology Unit of the HUS Children's Hospital
Prof. Kai Kaila – Laboratorof the University of Helsinki
Prof. Juha Voipio – Laboratorof the University of Helsinki
The project researchers at the Department of Signal Processing and Acoustics are Unto K. Laine, Okko Räsänen and Toomas Altosaar. Patents have been applied for these new pattern recognition methods.
For more information about the new pattern recognition methods, please contact:
Professor Unto K. Laine, unto.laine [at] aalto [dot] fi
and Researcher Okko Räsänen, okko.rasanen @ aalto.fi
Image: HYKS/Sampsa Vanhatalo