A Wavelet Transform and Neural Network Based Segmentation & Classification System For Bone Fracture Detection

Abstract

This paper deals with fracture detection of a human bone from X-ray images applying image processing techniques. It describes the wavelet transform based segmentation and neural network-based classification method for medical images, especially X-ray images, to locate or detect any fracture in the bones. Among different image segmentation techniques, using Wavelet Transform based segmentation manifests particular significance compared to other conventional segmentation techniques for the images considered in this study. Also in terms of statistical image identifier, namely Signal to Noise Ratio (SNR), Peak Signal to Noise Ratio (PSNR), Structural similarity (SSIM), and Entropy etc. The proposed methods have observable advantages, particularly in medical image segmentation. The Haar Wavelet showed maximum correlation with the considered images. The detailed vertical, horizontal and diagonal components can decompose X-ray images based on the Wavelet Transform algorithm. The vertical detail component of third level decomposition of the Wavelet Transform has preferred because it shows the minimum Entropy. The Error Backpropagation Neural Network (EBP-NN) fed with the obtained medical images from wavelet-based segmentation technique. This neural network has trained with fractured and non-fractured images then tested on various other X-ray images. An EBP-NN classification system with the architecture of 1024−22-2 gives the maximum accuracy. The developed system can detect fractured bone images accurately.