Text
Letting the Brain Help the Body — with AI and Brain Signals
Every year, millions of people suffer a stroke. Many are left with lasting motor impairments that make everyday life difficult — and in severe cases, impossible. Traditional rehabilitation helps, but for those who can barely move their hand, even participating in the training becomes challenging.
Can the brain's signals be used directly to support recovery, even when the body cannot?
The short answer is yes. With the help of brain–computer interfaces (BCI), patients can imagine movements and receive feedback based on their brain activity. This mental training activates motor networks and can stimulate the brain's plasticity. But for the technology to work in practice, the systems need to become faster, more reliable, and above all more adaptable to each user's brain.
The Core of the Research — understanding and improving how BCIs interpret the brain
The research focuses on three major challenges:
1. High-dimensional brain signals that must be understood in real time
EEG data contains vast amounts of information from many electrodes and frequencies. To provide feedback to the user, the system must quickly identify patterns corresponding to motor imagery — for example, imagining clenching one's hand.
Jonatan Tidare's work develops and evaluates methods that:
- classify motor imagery from one hand,
- use multivariate analyses to track signal strength over time,
- find relevant EEG features using genetic algorithms.
2. Requirements for real-time feedback
Feedback that arrives too late disrupts learning. The dissertation analyzes the system's total latency and shows which components cause bottlenecks — and how realistic thresholds for clinical use can be achieved.
3. Brain signals that change from day to day
EEG is notoriously unstable. A system that works excellently today may perform worse tomorrow.
Why This Matters
Today's clinical BCI systems are often static and lack the ability to adapt to patients' shifting abilities. This research shows that:
- adaptive BCIs can keep training meaningful,
- continuous measures of brain signal strength enable fine-tuned feedback,
- difficulty can be regulated automatically, similar to resistance on an exercise machine.
For stroke survivors, especially those with severe impairments, this can make the difference between being able to participate in rehabilitation — or not.
Results with Clinical and Technical Value
The dissertation contributes several practical and tested methods:
- time-resolved multivariate decoding that shows when and how strong a motor imagery signal is,
- a continuous metric that can be linked to visual representations and used to control difficulty in real time,
- a method for adaptive control of training level, tested in a clinical two-case study,
- strategies for handling EEG non-stationarity, confirmed across multiple training sessions.
In a case study with two stroke patients, the methods demonstrated that adaptive systems can provide more personally tailored and responsive rehabilitation, even when users have very weak or unstable signals.
Societal Benefit — a step toward future neurorehabilitation
The research strengthens the possibilities of offering:
- rehabilitation even for patients who cannot move,
- more effective and engaging training,
- more precise and individualized treatment protocols.
In a society where stroke is one of the leading causes of long-term disability, technologies that can enhance the brain's own recovery mechanisms are of great importance.
Relevance for MDU — Trusted Smart Systems
The dissertation aligns fully with MDU's focus on trusted smart systems. Here, the following converge:
- neuroscience,
- machine learning,
- time-critical real-time systems,
- and clinical needs.
Jonatan Tidare's work demonstrates how advanced, data-driven systems can be made robust enough for use with real patients — thereby contributing to MDU's ambition to develop technology that is both safe, adaptable, and of benefit to society.
Read more in Jonatan Tidare's doctoral thesis -> MDU Diva