Detection of bone tumours in radiographic images using neural networks

We develop an approach for segmenting radiographic images of focal bone les ions possibly caused by bone tumour. A neural network is used to classify i ndividual pixels by a convolution operation based on a feature vector. We d esign eight features chat characterise the local texture in the neighbourho od of a pixel. Four of the features are based on co occurrence matrices com puted from the neighbourhood. The true class label of the pixels in the rad iographs are obtained from annotations made by an experienced radiologist. Neural networks and self-organising feature maps are trained to perform the segmentation cask. The experiments confirm che feasibility of using a feat ure-based neural network for finding pathologic bone changes in radiographi c images. An analysis of the eight features indicates that the presence of edges and transitions, the complexity of the texture, as well as the amount of high frequencies in che texture, are che main features discriminating ( soft) tissue from pathologic bone, the two classes most likely to be confus ed.