Abstract:
A bone fracture is a complete or incomplete discontinuity in a bone, often caused by an impact. While extreme fractures are sometimes obvious, most fractures require radiographic imaging (such as X-ray) to diagnose and treat. Unfortunately, cost, access to such equipment, and availability of trained personnel to interpret the results present significant barriers to many in remote areas and developing countries. In this feasibility study, a low-cost and portable bone fracture detection method and device are proposed to help this under-served segment of patients. Drawing on previously published work regarding the automated detection of mechanical fractures using induced vibrations in an industrial setting, this paper presents a technique to replicate and improve upon manual detection techniques using a tuning fork and stethoscope by using digital signal processing and machine learning techniques. In order to make fracture detection more accessible, the prototype device presented does not require any specialized skills to operate, maintains portability, is automated, and has the potential to be manufactured inexpensively. Fractures are detected by inducing vibrations in the bone and measuring the resulting signal to detect structural defects. Using animal bones with synthetic soft tissues to replicate the dampening effects of muscle and connective tissue, machine learning models were trained and tested, achieving 93.6% accuracy. The proposed technique may also prove effective in-vivo although further testing is required.