نهمین کنگره نقشه برداری مغز

نمایش مقاله

Title:

Adaptive EEG Based Brain Computer Interface for Real-Time Control of a Surgical Robot in Simulated Tumor Excision

Subject:

Brain–Computer Interfaces (BCIs)

Presenter Author:
Name:	Morteza Shahedifar

Author:
Name 1:	Morteza Shahedifar - Corresponding - Presenter
Affiliation 1:	Electrical and Computer Eng. University of Tabriz, Iran

Name 2:	Alireza Siadatan
Affiliation 2:	Associate Professor, Senior Member of IEEE, University of Toronto, Canada

Name 3:	Mahdis Heidarnezhad
Affiliation 3:	Department of Engineering, Islamic Azad University, Tabriz, Iran

Name 4:
Affiliation 4:

Name 5:
Affiliation 5:

Name 6:
Affiliation 6:

Name 7:
Affiliation 7:

Name 8:
Affiliation 8:

Name 9:
Affiliation 9:

Name 10:
Affiliation 10:

Abstract

Original Research

introduction :

Precise control of surgical robots through noninvasive brain computer interfaces (BCIs) remains a critical challenge in neurosurgery and robotic assisted procedures. Converting EEG signals into accurate, real-time control commands requires methods that can adapt to variability in neural activity across individuals and over time. Traditional BCI approaches often rely on static classifiers and pre-defined mappings, which can limit responsiveness and precision, especially in tasks demanding fine motor control. Integrating adaptive deep learning frameworks allows the system to continuously recalibrate based on the user’s EEG patterns, potentially enabling robust and stable control of robotic surgical tools. Simulation-based evaluation provides a safe environment to test these models, quantify performance, and refine control strategies before any in vivo application.

Methods:

EEG signals were recorded from 20 participants performing motor imagery tasks corresponding to key robotic movements, such as reach, grasp, and tool manipulation. Preprocessing included band-pass filtering in motor-relevant frequency bands, artifact rejection, and normalization. A compact adaptive deep learning model was trained to map EEG features to control commands for a simulated RAVEN II surgical robot executing tumor excision tasks. The architecture incorporated online adaptation mechanisms to handle temporal drift and inter-subject variability. Real-time feedback loops were implemented in simulation to adjust movement trajectories and ensure smooth, safe operation. Multiple simulation scenarios were explored, including sequential excision motions, tool orientation adjustments, and sudden changes in intended motion. Performance metrics focused on trajectory accuracy, execution time, and movement smoothness.

Results:

Simulations showed that the adaptive BCI successfully translated EEG signals into precise robotic movements, achieving low trajectory error and minimal latency across participants. Online adaptation reduced abrupt corrections and improved smoothness compared to non-adaptive baselines, particularly under signal variability. Complex sequences, including multi-joint tool adjustments, were executed reliably in simulation, demonstrating that EEG-based commands can drive realistic surgical maneuvers. Analysis of control signals highlighted that adaptive learning maintained consistent mapping from neural activity to robotic motion, even when individual EEG patterns shifted over time. These results suggest that the system is robust to inter-subject differences and can maintain performance in tasks requiring fine precision.

Conclusion:

Combining adaptive deep learning with EEG-based BCI enables real-time control of surgical robots in simulated tumor excision tasks. By focusing on neural variability, continuous adaptation, and simulation-based evaluation, the framework demonstrates both precision and robustness in complex robotic movements. This approach provides a reproducible and safe foundation for advancing EEG-driven robotic surgery and highlights the potential of noninvasive BCI systems in high-precision medical applications.

Keywords:

BCI; EEG; Adaptive Learning; Robotic Surgery; Motor Imagery