Beyond performance metrics: dynamic directed functional connectivity as neural biomarker of surgical skill proficiency
Loading...
Authors
Kamat, Anil, Kumar
Issue Date
2025-05
Type
Electronic thesis
Thesis
Thesis
Language
en_US
Keywords
Mechanical engineering
Alternative Title
Abstract
Objective assessment of surgical skills is critical for ensuring proficiency, enhancing professional certification processes, and improving patient safety. Existing methods, however, rely on subjective human judgment, introducing bias and limiting reproducibility. While recent approaches have leveraged kinematic data and neural imaging to provide more objective evaluations, these methods evaluate performance outcomes, functioning as metrics, i.e., output measures that help explain what happened during a performance, such as how quickly or how accurately the task was completed. While metrics are valuable for providing a snapshot of performance, they fall short of explaining the underlying neurophysiological process driving the performance, i.e. how and why skill acquisition occurs at the neural level.This thesis introduces dynamic directed functional connectivity (dFC) as a novel neural biomarker that has potential for surgical skill assessment. Unlike traditional metrics, biomarkers offer a physiological basis for evaluation, capturing both the strength and direction of neural information flow in the brain to provide deeper insights into the brain's role in motor skill acquisition and proficiency. Using electroencephalography (EEG) and an attention-based Long Short-Term Memory (LSTM) model to compute non-linear Granger causality, we quantify dFC among key brain regions involved in psychomotor surgical task execution. As a biomarker, dFC enables the evaluation of dynamic neural processes underlying learning and skill execution at both group and individual levels, complementing existing performance metrics. Coupled with hierarchical task analysis (HTA), dFC facilitates subtask-level assessments, while a convolutional neural network (CNN) classifies skill levels with higher accuracy and specificity than traditional approaches in laparoscopic surgery.
For the first time, dFC is applied to map stages of the well-established Fitts and Posner motor learning model, providing new insights into the neural mechanisms underlying skill acquisition and retention. dFC effectively identifies and tracks progression through various stages of this model, and its stability over a six-week washout period highlights its utility in monitoring long-term retention. Importantly, control group analysis confirms that observed neural adaptations are specific to training rather than external factors.
Additionally, we investigate the neurophysiological impact of physical versus virtual reality (VR) based simulators on brain connectivity during surgical tasks. Functional near-infrared spectroscopy (fNIRS) combined with Granger causality revealed a statistically significant effect of training on physical and virtual simulators, which depends on expertise level. Short-separation channel analysis ensures that these effects reflect neurophysiological changes rather than extracerebral artifacts.
By establishing dFC as a robust, biologically grounded, objective, and reproducible neural biomarker, this work provides a comprehensive and individualized framework for evaluating and optimizing surgical training protocols. This work addresses critical limitations of subjective and performance-based metrics, paving the way for personalized training protocols, improved simulator designs, and a step towards developing evidence-based certification standards. Ultimately, these advancements promise to elevate surgical education, enhance training efficiency, and bolster patient safety.
Description
May2025
School of Engineering
School of Engineering
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY