Directed functional connectivity outperforms score-based tracking in laparoscopic training
A brain-based signal tracked laparoscopic motor learning more cleanly than a standard performance score. In a 15-day suturing program, directed functional connectivity derived from electroencephalography separated pre-training from post-training trials with 94.4% accuracy, exceeding the Fundamentals of Laparoscopic Surgery-based classifier at 84.5%.
The study focused on a problem that surgical education has long faced at the bedside and in the skills lab: performance scores can improve while the neural machinery of learning remains hidden. By measuring information flow across five cortical regions with nonlinear Granger causality, investigators built a biomarker that aimed to move beyond whether a trainee simply completed a task and toward how the brain reorganized to support the task. That distinction matters in procedural training, where durable competence depends not just on repetition but on consolidation of the motor program.
Twenty-five trainees completed the training program and were tested again six weeks later. An untrained control group of 27 participants and 15 expert surgeons served as behavioral and autonomous-stage references. The design allowed the biomarker to be examined across the full arc of skill acquisition, from early practice through retention, while also comparing trainee trajectories with two clinically meaningful anchors: lack of practice and established expertise.
Neural adaptation moved on a different clock than behavior
A sharp separation emerged between neural and behavioral timing during the shift from novice to more automatized performance. Directed functional connectivity marked the cognitive-to-associative transition on day 6, one day later than the behavioral metric, which identified the change on day 5.
Sleep-dependent consolidation offers a plausible explanation for that lag. Behavioral improvement can appear first at the bench, while cortical networks continue refining how information flows between regions after practice ends. In surgical training, that sequence has practical consequences: a trainee may look competent earlier than the underlying network has fully stabilized, and a biomarker that captures the brain’s slower adaptation could expose that mismatch.
Retention testing added another layer. For pre-training versus retention discrimination, the EEG-based classifier and the FLS-based classifier performed closely, at 87.5% and 88.5% respectively. After the immediate post-training phase, neither measure changed further from post-training to retention, and the control group showed no adaptation. The retention result suggests that once the neural and behavioral systems reached a post-training plateau, the biomarker no longer registered meaningful drift over the six-week interval.
One sentence captures the central result: the brain signal detected learning with greater precision than the score tied to task performance.
Why a network signal may be more informative than a single score
The appeal of directed connectivity is biological as much as statistical. Surgical learning is not a one-node event; it depends on coordinated communication among cortical regions that support attention, planning, sensorimotor integration, and error correction. A unilateral score, even a well-validated one like the FLS, compresses that complexity into a single output. Directed connectivity instead interrogates the pattern of influence between regions, making it better suited to detect the reconfiguration that accompanies skill acquisition.
That network-level view also explains why the biomarker tracked subtask-level changes. The investigators used subtask decomposition, change-point detection, and a one-dimensional convolutional neural network benchmarked against the FLS score. In practice, this means the signal was not limited to a broad pre/post contrast; it could follow smaller inflection points within the training curve and align them with behavior over time. Individual-level dFC measures also tracked behavioral cumulative sum trajectories, suggesting that the neural marker and the performance curve were moving in parallel even when their exact timing differed.
Experts served as the autonomous-stage reference, an important choice because mastery in surgery is not simply the ability to repeat a motion. It is the ability to execute the motion with reduced cognitive load, greater efficiency, and less dependence on deliberate control. A biomarker that maps onto that transition could eventually help trainers decide when a learner has crossed from effortful practice into a more stable motor repertoire.
Training programs could gain a second endpoint
The immediate implication is not replacement of behavioral assessment but augmentation. FLS scores remain useful because they are practical, familiar, and easy to interpret. Yet a biomarker that identifies learning-related reorganization in the cortex could provide a second endpoint when performance alone is ambiguous, especially in complex procedural tasks where two trainees may earn similar scores while using very different neural strategies.
For surgical educators, the attraction lies in longitudinal monitoring. A stable biomarker could flag when a trainee is still consolidating, when the learning transition has shifted, and when the network has largely plateaued. That could matter for curriculum design, remediation, and decisions about whether more deliberate practice is still needed. It could also help researchers compare simulation platforms not only by the speed of score improvement but by the quality of neural adaptation they induce.
The data support that ambition, but they do not yet establish a clinical deployment standard. The sample was modest, the setting was a focused laparoscopic suturing program, and the work was validated against a specific benchmark task rather than a broad surgical curriculum. Even so, the result is unusually concrete: EEG-based directed connectivity outperformed a score-based model for distinguishing pre- and post-training states and preserved strong discrimination at retention.
A biomarker still needs wider surgical validation
The next question is whether the same approach will hold across different procedures, skill levels, and training environments. Laparoscopic suturing is demanding, but surgery is not a single motor domain. Open, endoscopic, and robotic tasks likely engage overlapping yet distinct neural circuits, and a biomarker that succeeds in one setting may need recalibration in another. The article’s own framing makes clear that the promise lies in brain-based biomarker development, not in claiming that one EEG model can already replace entrenched assessment systems.
Funding came from the Medical Technology Enterprise Consortium and the U.S. Army Futures Command, Combat Capabilities Development Command Soldier Center. Those sources fit the translational nature of the work: the intended use is not abstract neuroscience, but performance monitoring for real-world skill acquisition. The authors declared no competing interests.
For now, the strongest conclusion is narrow and useful. Directed connectivity captured learning-related neural adaptation with finer resolution than a conventional score, reached the cognitive-to-associative transition one day later than behavior, and remained stable through retention without further change in controls or trainees. The remaining task is to determine whether that network signature can travel beyond one training model and become a dependable adjunct for surgical education.