AI and infra-red open communication with ‘locked-in’ people

Brain researchers have used artificial intelligence and infra-red sensing to deduce yes/no responses from four paralysed people.

near infra-red brain interface Wyss Center

This is not the first time communication from completely paralysed people has been enabled using some form of brain interface, but it is an early example of communication via natural yes/no thoughts – rather than through the subject’s training their thoughts to match the needs of the interface. A related team of researchers deduced yes/no answers from infra-red sensing of a single person in 2014.

Attempts to do something similar using electrical electroencephalogram (EEG) sensing have been unsuccessful, according to ‘Brain-computer interface-based communication in the completely locked-in state‘, a paper in PLOS Biology which describes the new work.

The scanning technique, dubbed ‘functional near-infrared spectroscopy’ (fNIRS) is non-invasive, and involves infra-red sources and sensors places on the scalp. These are used to detect changes in oxygenated blood flow in the brain – flows known to change during thinking, in this case ‘frontocentral’ oxygenation changes.

The paralysed people were asked many questions with obvious yes or no (or true or false) answers, with the correct answers plus the scanned data fed as training data into an artificial intelligence scheme known as a ‘linear support vector machine’ which is can be used for classification tasks – deciding between ‘yes’ or ‘no’, for example.

Once trained, the artificial intelligence delivered yes or no responses on new questions with around 70% accuracy.

Inaccuracy was not spread evenly and, according to the paper, in some cases corresponded to certain EEG patterns “probably because of decreased vigilance and attention”.

In addition to functional near-infrared spectroscopy, methods tried to create brain-computer interfaces include:

in-brain electrodes
brain surface electrodes
electroencephalography
functional MRI

Organisation involved in this research include: University of Tübingen, Shanghai Maritime University, Central Institute of Mental Health (Mannheim, Gremany), National Institute of Neurological Disorders and Stroke (Maryland), Ospedale San Camillo (Venice) and the Wyss Center (Geneva).

The photo above is from the Wyss Center.