Neuroimaging biomarkers for surgical skill level prediction

We have established in previous studies that functional near-infrared spectroscopy (fNIRS) enable to retrospectively classify subject surgical skills levels when performing the Fundamentals of Laparoscopic Surgery (FLS) program tasks. In this work, we further investigate the utility of fNIRS to complement current surgical assessment methodologies by demonstrating that the combination of fNIRS with Machine Learning allows predicting the FLS scores that are currently employed in surgical certification. Technically, the oxy- and deoxyhemoglobin concentration time course data from prefrontal cortex (PFC), supplementary motor area (SMA), and primary motor cortex (M1) were acquired while the subjects were performing the FLS pattern cutting task. Machine learning models leveraging seven fNIRS derived biomarkers were trained to predict the FLS scores. The results obtained confirm that our methodology is able to accurately predict FLS scores based on neuroimaging data, as independently assessed via R2 values. Hence, these results establish that fNIRS is potentially well suited for bed-side, real time and cost effective assessment of bimanual skill levels for surgical certification.