When brain cells act together, they produce brain waves known as oscillations, which can be measured using certain types of brain scanner. Analysing the features of these oscillations reveals clues about how the brain is working, and there is even evidence that they differ in people with some forms of dementia, including Alzheimer’s disease.
Most existing tools to analyse brain oscillations assume the waves are neat and even, known as sinusoidal waveforms. However, in the active brain, most oscillations are non-sinusoidal and are made up of lots of different wave patterns. This makes the existing analytical tools less accurate when measuring complex non-sinusoidal brain waves.
Now, PhD student Marco Fabus, who works in the Oxford Centre for Human Brain Activity, has devised a tool to analyse these non-sinusoidal waves, which is detailed in a new paper published in the Journal of Neurophysiology. His method – called iterated masking empirical mode decomposition (itEMD) – involves capturing the maximum points of the waves to isolate the highest frequencies. This process is then repeated for the next highest frequency points until all the individual sub waves have been separated from the original wave. It further improves on existing methods by introducing helper signals which are determined in a data-driven way.
Marco Fabus said: ‘My background is in physics, so I apply physics principles to neuroscience to look at it in a new light – it’s how I came up with the iterated method. I am passionate about the benefit of interdisciplinary connections. Within our research team, we have researchers with undergraduate degrees ranging from psychology to engineering, and everyone contributes something different.’
This process of repetition – or iteration – is what makes this new analysis technique so robust. It is also largely automatic, requiring less input from those using it. But the important thing about itEMD is that it doesn’t make any assumptions about the waveform shape (unlike most existing tools), which makes it far more useful.
Largely thanks to the efforts of DPUK researcher Dr Andrew Quinn, itEMD is now available on Python, a software commonly used by researchers to analyse their data. Marco Fabus said: ‘Open science is very important to me. We have made our entire methodology publicly available, so any researcher could exactly reproduce our data. Plus, now itEMD is on Python, it is available for use by researchers in any field, increasing the ramifications of our work.
One of the ways itEMD will be used is by the team in DPUK’s New Therapeutics in Alzheimer’s Disease (NTAD) study to identify early biomarkers of dementia. NTAD is a collaboration between university and industry scientists that aims to detect these subtle biomarkers, allowing researchers to test more quickly whether a new treatment is likely to work. Slight differences in the oscillations of brain waves between people with and without dementia – only visible thanks to the precision of this new analytical tool – could indicate initial signs of the condition and allow earlier detection and intervention.
Dr Andrew Quinn said: ‘Often, we want to investigate detailed features of these oscillations, but this is challenging using conventional analysis methodologies that need long time-windows to work and smooth across the details. Marco’s paper develops and validates a method that will be an important part of NTAD analyses once baseline data collection is complete later this year.’
Another potential implication of the research is not just measuring brain waves but altering them with electric or magnetic stimulation. In this emerging area of dementia treatment, there is tentative evidence that stimulating the brain at specific points of an oscillation could improve cognition – and with itEMD, it may be possible to identify these peaks for targeted stimulation.
Marco Fabus and his colleagues are now building on this work with a new paper, currently available on the preprint server bioRxiv. This latest work explores non-sinusoidal waveforms in more detail and aims to categorise smaller waves into either distinct oscillations or subcomponents of larger, non-sinusoidal waves.