Bone fracture healing assessment using a neural network

Abstract

A method is proposed to estimate the strength of a healing fractured bone. A four electrical network model of the low frequency (50Hz-1 kHz) transverse vibration response taly. port of a healing fractured bone is derived using Timoshenko beam theory. A back-propagation neural net with one hidden layer processes computer simulated bone fracture model data, and classifies it with respect to the fracture gap stiffness, relative to intact bone. The effect on classification accuracy of number of hidden units and preprocessing of the data is evaluated. Vibration measurements from a rabbit fracture model and clinical patient data are also presented.