Thresholding Technique for Accurate Analysis of Density and Geometry in QCT, pQCT and μCT Images
Computed tomography (CT) is widely used in the assessment of bone parameters in live patients and animals as well as bone samples. Quantitative analysis requires the segmentation of the bone from the surrounding tissue, and most segmentation methods rely on some type of thresholding technique. The aim of this communication is to highlight the influence of threshold selection on various bone parameters and recommend appropriate thresholds. Two types of information are of interest in bone analysis from images: geometric parameters and density parameters. We know from imaging theory that blurring is an inherent byproduct of all imaging methods. Depending on the threshold used for segmentation, the object boundary moves in space due to the sloping edge. It is, thus, critical to select the threshold that creates an object boundary that reflects the actual object size. Similarly, due to blurring, the imaged density shows erroneous values at the object boundaries. Such values must not be included for an accurate representation of the object density. Using a pQCT scanner and a bone phantom with known density and geometry, we show that the thresholds for geometry and density are different. The threshold for accurate geometric segmentation was 49% of the difference of the density between the adjacent tissues. The threshold for accurate density assessment was 95% of the maximum density value of the bone. These specific thresholds are valid only for the scanner tested; however, the principle for selecting the thresholds is valid across scanner platforms and scale of imaging.
Hangartner, T. N.
(2007). Thresholding Technique for Accurate Analysis of Density and Geometry in QCT, pQCT and μCT Images. Journal of Musculoskeletal and Neuronal Interactions, 7 (1), 9-16.