Cancellous bone structure analysis using image analysis

The application of image analysis techniques for the analysis of histology are developing rapidly particularly where it is possible to prepare high contrast histology sections. Bone histology is one such instance, bone matrix can be stained with van Geisen. This preparation is particularly suitable for image segmentation using grey level threshold detection. This study has quantified the bias and random error that can be attributed to pixel sizing and the threshold detection level. Also, the influence of the working magnification and the intra and inter observer variability on the reproducibility and reliability of image analysis results. The study has analysed vertebral cancellous bone structure in vertebrae with low bone volume (5.53%) and high bone volume (20.85%) at a range of magnification (x4 to x640). Data were collected at various machine settings and at different times by the machine operators. Data were analysed using Student's paired t-test. Variation from the optimum threshold detection level significantly influences the magnitude of bone volume, bone mineral surface and trabecular thickness measurements. Bone mineral surface measurement increases as the working magnification increases, identifying the fractal nature of cancellous bone. Operator bias is less than 5% though random error can be as high as 13%. Image analysis can be very efficient and make quantitative tissue analyses more accessible to scientists interested in studies of comparative pathology. This study shows that reproducible and reliable data for the analysis of cancellous bone structure can be obtained by the application of a clearly defined operating protocol and adequate operator training.