Neural Network Analysis of Bone Vibration Signals to Assesses Bone Density

Osteoporosis is a systemic disease, characterised by low bone mineral density (BMD) with a consequent increase in bone fragility. The most commonly used method to examine BMD is dual energy X-ray absorptiometry (DXA). However DXA cannot be used reliably in children less than 5 years old because of the limitations in the availability of required normative data. Vibration analysis is a well-established technique for analysing physical properties of materials and so it has the potential for assessing BMD. The overall purpose of this study was development and evaluation of low frequency vibration analysis as a tool to assess BMD in children. A novel portable computer-controlled system that suitably vibrated the bone, acquired, stored, displayed and analysed the resulting bone vibration responses was developed and its performance was investigated by comparing it with DXA-derived BMD values in children. 41 children aged between 7 and 15 years suspected of having abnormal BMD were enrolled. The ulna was chosen for all tests due to the ease with which it could be vibrated and responses measured. Frequency spectra of bone vibration responses were obtained using both impulse and continuous methods and these plus the participants’ clinical data were processed by a multilayer perceptron (MLP) artificial neural network. The correlation coefficient values between MLP outputs and DXA-derived BMD values were 0.79 and 0.86 for impulse and continuous vibration methods respectively. It was demonstrated that vibration analysis has potential for assessing fracture risk