Application of machine learning in rotary ultrasonic-assisted orthopedic bone drilling: A biomechanical pull out in vitro study

Abstract

Bone drilling is a mechanical, thermal coupling process utilized in orthopedics for provision of rigid internal fixation and treatment of fractured bone. Rotary ultrasonic-assisted bone drilling (RUABD) has achieved noteworthy interest in orthopedic practice due to its ability to enhance biomechanical pullout strength. Drilling parameters used during orthopedic surgeries significantly impact the holding power, initial implant stability and pullout strength. It is difficult for surgeons to predict push-out strength at the interface of bone and screw. An intelligent approach could involve utilizing machine learning (ML) to train and test independent drilling parameters, thereby predicting pullout strength and optimizing holding strength. Therefore, the present work focused on leveraging ML models during RUABD to predict pullout strength at the bone-screw interface. The monitoring of various drilling parameters (including insertion angle, feedrate, rotational speed, and ultrasonic amplitude) was conducted. Multiple ML models were employed to forecast the pullout strength at the interface between bone and screw. The SVR-based ML model exhibited the most accurate prediction among all models, with the lowest error metrics observed. ML algorithms can be leveraged for robust prediction of biomechanical pullout strength to upsurge holding strength and avoid screw loosening.