Deland Liu Successfully Defended his PhD Thesis. Congratulations!!

October 2025, Deland Liu successfully defended his Phd thesis:

 “Error-related Potentials in Neural Engineering: From Perceptual Skill Enhancement to Assistive Communication”

Let’s Party!

We wish you all the best in your future career

 

Detailed Thesis Abstract:

Communication and error monitoring are fundamental to human interaction, motor control and learning. Detecting and correcting errors allows us to maintain motor precision and adapt through experience, while communication enables social interaction and independence. Neural engineering seeks to support or enhance these functions by recording, decoding, and modulating brain activity, often through electroencephalography-based brain–computer interfaces (EEG-BCIs). Within this context, the error-related potential (ErrP)—an event-related potential elicited by the perception of an error—emerges as a particularly relevant neural signal. ErrPs serve both as biomarkers of error perception and candidate control signals for EEG-BCI communication. This dissertation expands the role of ErrPs in neural engineering by addressing two central challenges.

The first part of the dissertation investigates whether ErrPs can be leveraged to enhance perceptual sensitivity in healthy individuals. Accurate error perception is critical for maintaining motor precision in high-stakes settings such as surgery and sports, yet conventional behavioral training often fails to improve sensitivity to subtle errors. Chapter 2 introduces a closed-loop BCI that delivers real-time feedback on 7ErrP presence during perceptual training. This intervention enhanced sensitivity to subtle visuo-motor errors, overcoming performance bottlenecks seen in conventional, behavioral training. Perceptual gains were accompanied by increases in Pe amplitude, a hypothesized neural correlate of error perception. Chapter 3 extends this approach using personalized theta-band transcranial alternating current stimulation (tACS) applied over the frontocentral region. A single 20-minute session amplified Pe amplitude relative to control and enhanced error perception. This provides the first evidence that brain stimulation can augment the Pe component of the ErrP and, in turn, improve error perception.

The second part of the dissertation examines ErrPs as potential control signals in BCIs for restoring communication in motor-impaired patients, while also exploring alternative EEG control signals and complementary AI frameworks that could eventually integrate with ErrP-based BCIs to restore communications. Amyotrophic lateral sclerosis (ALS) is a devastating neurodegenerative disease that can progress to the completely locked-in state (CLIS), where patients lose all voluntary muscle function and, with it, conventional means of communication. Noninvasive EEG-BCIs are promising for restoring autonomy and communication in CLIS, uing ErrP as a candidate control signal. However, evidence of ErrP preservation in CLIS is scarce, and their reliable online decoding has not been demonstrated. Chapter 4 presents the first evidence that ErrPs are preserved in an ALS patient in CLIS, showing valid temporal, spectral, and spatial signatures. However, single-trial decoding was unreliable, limiting their immediate use as control signals. To address the urgent need for communication, we developed an alternative EEG-based BCI that relied on volitional modulation of alpha- and beta-band power, paired with auditory feedback to accommodate visual impairments. Using this system, the CLIS patient was able to respond to general knowledge and personally relevant assistive needs across multiple sessions, offering rare evidence that noninvasive EEG-BCIs can enable functional communication in CLIS. Chapter 5 then introduces a proof-of-concept framework that integrates large language models (LLMs) with binary neural input to achieve intent-level communication, shifting away from conventional symbol-by-symbol spellers. In a healthy participant using electromyography (EMG) and under simulated ErrP decoding accuracies, the system inferred naturalistic communicative intents with relatively high throughput and low number of decoding steps.

In summary, this dissertation advances the scientific understanding and applications of ErrPs in neural engineering. As a biomarker, ErrPs can be targeted to enhance perceptual sensitivity in healthy individuals through closed-loop BCIs and brain stimulation. As potential control signals, ErrPs are preserved in CLIS but remain difficult to decode reliably. In the meantime, alternative EEG signals provide practical routes to restore communication, while LLM-based frameworks demonstrate how ErrP-based BCIs could one day be extended to enable richer, intent-level communication.