Textons for 3D Binary Data with Applications to Classifying Cancellous Bone

Abstract

Changes to bone density (BV/TV) due to ageing or diseases such as osteoporosis is well documented. Changes in the structure of cancellous bone are less well understood. In addition, changes in structure as a function of distance from the growth plate have not received much attention. One obstacle in studying structure changes is in quantifying the irregular shapes of trabeculae that constitute cancellous bone. Here, the method of texons, originally developed for texture analysis in images, is adapted to characterise patterns in three-dimensional binary data and used to characterise the structure of cancellous bone. Analysis of micro-CT scans of tibias from 30 growing rats in three experimental groups indicate that texton based structure characteristics are able to distinguish structure of cancellous bone both in the different treatment groups and as a function of distance from the epiphyseal growth plate.