A Study of Vibration Signal Feature Extraction for Bone Drilling Layer Classification

In modern orthopedic surgery, bone drilling is an essential and popular technique used for various purposes such as fracture fixation, bone grafting, joint arthroplasty, and others. This study focuses on measuring and analyzing drilling vibration in orthopedic surgery during bone drilling to address common issues that may arise. The vibration signal of the drill is measured using accelerometer sensors placed near the drill and on the bone in order to avoid any unnecessary vibration signal attenuation. This research specifically examines bone drilling on a cow femur as its bone composition is similar to that of humans. 50,000 vibration signal data are obtained and will be researched to feature extraction for the purpose of classifying layers in bones. The feature extraction method (also called characteristics and signatures) that will be used in this study is presented and grouped into two categories: time-domain features extraction and frequency-domain features extraction. The data collected on the x-, y-, and z-axis of ten drilled holes are used to get feature extraction which is grouped into three classes: layer 1 (periosteum), layer 2 (first cortical), and layer 3 (spongy). The study identifies six feature combinations that have feature values that do not overlap with each other. In the time-domain feature extraction, Mean, RMS, Skewness, and Kurtosis are extracted, while in the frequency-domain feature extraction, root mean square frequency (RMSF), and root variance frequency (RVF) produce good classification results compared to other features.