Parametric Power Spectral Density Estimation-Based Breakthrough Detection for Orthopedic Bone Drilling with Acoustic Emission Signal Analysis

Manual bone drilling in orthopedic surgical operations may cause injury to patient tissues if the drill bit continues to progress after exiting the bone. In this study, a new bone breakthrough detection algorithm based on acoustic emission (AE) signal analysis has been developed to minimize temporary and permanent injuries that can be caused by surgeon-controlled surgical drills. Three parametric estimation methods, Burg, Yule-Walker and Modified Covariance were used to estimate Power Spectral Density (PSD) of the AE signal during the drilling operation. Four frequency features, Mean Frequency, Median Frequency, Mean-Median and Power Bandwidth were calculated for each PSD estimate. An artificial neural network-based breakthrough detection classification was constructed from the extracted features. The highest breakthrough detection performance was obtained with the features extracted by the Burg method with an accuracy rate of 90.95 +/- 0.97% in the training phase and 92.37 +/- 1.09% in the test phase. In the detection of Not-Breakthrough situations, the highest accuracy was obtained with features extracted with the Covariance method as 99.04 +/- 0.03% in the training phase and 99.05 +/- 0.08% in the testing phase. This new approach which could be integrated into conventional drills with minimum configuration changes and without any major cost has the potential to increase the performance and safety of bone drilling procedures.