Sensor and machine learning-based assessment of gap balancing in cadaveric unicompartmental knee arthroplasty surgical training.

ABSTRACT

PURPOSE: The aim of this study was to assess the difference between flexion and extension contact forces-gap balance-after Oxford mobile-bearing medial unicompartmental knee arthroplasty (UKA) performed by surgeons with varying levels of experience. METHODS: Surgeons in a training programme performed UKAs on fresh frozen cadaveric specimens (n = 60). Contact force in the medial compartment of the knee was measured after UKA during extension and flexion using a force sensor, and values were clustered using an unsupervised machine learning (k-means algorithm). Univariate analysis was performed with general linear regression models to identify the explanatory variable. RESULTS: The level of experience was predictive of gap balance; surgeons were clustered into beginner, mid-level and experienced groups. Experienced surgeons' mean difference between flexion and extension contact force was 83 N, which was significantly lower (p < 0.05) than that achieved by mid-level (215 N) or beginner (346 N) surgeons. CONCLUSION: We found that the lowest mean difference between flexion and extension contact force after UKA was 83 N, which was achieved by surgeons with the most experience; this value can be considered the optimal value. Beginner and mid-level surgeons achieved values that were significantly lower. This study also demonstrates that machine learning can be used in combination with sensor technology for improving gap balancing judgement in UKA.