Technique for bone volume measurement from human femur head samples by classification of micro-CT image histograms

Micro-CT analysis is a powerful technique for a non-invasive evaluation of the morphometric parameters of trabecular bone samples. This elaboration requires a previous binarization of the images. A problem which arises from the binarization process is the partial volume artifact. Voxels at the external surface of the sample can contain both bone and air so thresholding operates an incorrect estimation of volume occupied by the two materials.

The aim of this study is the extraction of bone volumetric information directly from the image histograms, by fitting them with a suitable set of functions.

Nineteen trabecular bone samples were extracted from femoral heads of eight patients subject to a hip arthroplasty surgery. Trabecular bone samples were acquired using micro-CT Scanner. Hystograms of the acquired images were computed and fitted by Gaussian-like functions accounting for: a) gray levels produced by the bone x-ray absorption, b) the portions of the image occupied by air and c) voxels that contain a mixture of bone and air. This latter contribution can be considered such as an estimation of the partial volume effect.

The comparison of the proposed technique to the bone volumes measured by a reference instrument such as by a helium pycnometer show the method as a good way for an accurate bone volume calculation of trabecular bone samples.