Events

Public defence, Neuroscience and Biomedical Engineering, MSc Dovilė Kurmanavičiūtė

Intuitive Auditory Brain–Computer Interface.

Public defence from the Aalto University School of Science, Department of Neuroscience and Biomedical Engineering.
Doctoral hat floating above a speaker's podium with a microphone.

In our events, we are committed to Aalto University’s principles for a safer space.

Title of the thesis: Intuitive Auditory Brain–Computer Interface

Thesis defender: Dovilė Kurmanavičiūtė
Opponent: Research scientist Dr. Marie-Constance Corsi, Paris Brain Institute (ICM) & Inria Paris, France
Custos: Prof. Lauri Parkkonen, Aalto University School of Science

Brain–computer interfaces (BCIs) can provide a potential route to restore basic communication by decoding intention directly from brain activity, bypassing damaged motor pathways.

This thesis aimed to develop an intuitive, non-invasive auditory BCI for binary “Yes”/“No” communication to support patients with severe motor disabilities. The thesis comprised three research studies, of which the first investigated whether selective auditory attention to naturalistic, intuitive stimuli (spoken words “Yes” and “No”) could be decoded from unaveraged brain responses. This study utilized non-invasive magnetoencephalography (MEG) to capture the fine spatiotemporal brain signal changes related to attention. The results demonstrated high classification accuracy, with the most discriminative information coming from the auditory cortices bilaterally, establishing that intuitive verbal stimuli are viable for high-accuracy BCI control.

While MEG provides a high signal-to-noise ratio, its lack of portability limits BCI applications to research laboratories. The second study aimed to bridge the gap between laboratory MEG and portable electroencephalography (EEG). By systematically reducing the EEG channel count, this study mapped the trade-off between measurement complexity and decoding performance. The findings revealed that even with a significant reduction in sensors—down to nine EEG channels—usable classification accuracy could be maintained, provided the sensors were optimally placed. However, this study also identified a critical challenge: when classifiers were trained on data from the start of a session and tested on later data, accuracy dropped dramatically due to the non-stationarity of brain signals over time.

The final study addressed this performance degradation. To simulate real-world clinical deployment, a causal testing framework was employed where the system predicted user intent chronologically without access to future data. Under these realistic conditions, static classifiers performed poorly. To solve this, an adaptive algorithm was implemented to recalibrate the BCI trial-by-trial. This approach significantly improved communication speed, increasing the Information Transfer Rate (ITR) almost 10-fold. Furthermore, the work demonstrated that the system can be initialized using a general subject-independent model, enabling immediate operation before subject-specific adaptation.

Overall, this doctoral research demonstrates that while the auditory cortex provides discernible signals for attended versus non-attended “Yes”/“No” stimuli, the primary challenge for an intuitive BCI is the variability of those signals over time. By moving from the ideal conditions of MEG to the practical constraints of low-channel EEG and adaptive classification, this thesis provides a roadmap for transitioning BCI research from the laboratory to practical clinical use. The results highlight that a practical, intuitive BCI must employ a continuously adapting classifier capable of tracking the user's changing brain signals.

Key words: MEG, EEG, brain–computer interface (BCI), recursive least squares (RLS), machine learning, auditory responses, selective attention

Thesis available for public display 7 days prior to the defence at Aalto University's public display page. 

Contact information: dovile.kurmanaviciute@aalto.fi

Doctoral theses of the School of Science

A large white 'A!' sculpture on the rooftop of the Undergraduate centre. A large tree and other buildings in the background.

Doctoral theses of the School of Science at Aaltodoc (external link)

Doctoral theses of the School of Science are available in the open access repository maintained by Aalto, Aaltodoc.

Zoom Quick Guide
  • Updated:
  • Published:
Share
URL copied!