In the present paper we addressed the problem of whether the information about the structural anisotropy of trabecular bone can be retrieved from low-quality data, captured with clinical multi-raw spiral CT scanners. Two measures of quantifying structural anisotropy were tested - the current standard mean intercept length (MIL) and the gray-level structure tensor (GST). Thirty two vertebral bodies were μCT and CT scanned. The reference values of structural anisotropy were measured in μCT images and compared with the measures of structural anisotropy determined from low-quality CT data. MIL-based measures of structural anisotropy cannot be reliably determined from CT data. The assessment of the GST is significantly better than that of MIL, but the accuracy is not, in general, satisfactory. Based on the results of experiments with artificial data and the analysis of the real images, it can be concluded that a possible reason of the poor performance is anisotropic resolution of clinical CT scanners.

Rheumatoid arthritis (RA) is a disease characterized by progressive and irreversible destruction of bones and joints. According to current recommendations, magnetic resonance imaging (MRI) is used to asses three main signs of RA based on manual evaluation of MR images: synovitis, bone edema and bone erosions. The key feature of a future computer-assisted diagnostic system for evaluation RA lesions is accurate segmentation of 15 wrist bones. In the present paper, we focus on developing a wrist bones segmentation framework. The segmentation procedure consisted of three stages: segmentation of the distal parts of ulna and radius, segmentation of the proximal parts of metacarpal bones and segmentation of carpal bones. At every stage, markers of bones were determined first, using an atlas-based approach. Then, given markers of bones and a marker of background, a watershed from markers algorithm was applied to find the final segmentation. The MR data for 37 cases were analyzed. The automated segmentation results were compared with gold-standard manual segmentations using a few well-established metrics: area under ROC curve AUC, mean similarity MS and mean absolute distance MAD. The mean (standard deviation) values of AUC, MS and MAD were 0.97 (0.04), 0.93 (0.09) and 1.23 (0.28), respectively. The results of the present study demonstrate that automated segmentation of wrist bones is feasible. The proposed algorithm can be the first stage for the detection of early lesions like bone edema or synovitis.

The aim of the study was to create the efficient tool for semi-automated detection of bone marrow oedema lesions in patients with axial spondyloarthritis (axSpA). MRI examinations of 22 sacroiliac joints of patients with confirmed axSpA-related sacroiliitis (median SPARCC score: 14 points) were included into the study. Design of our algorithm is based on Maksymowych et al. evaluation method and consists of the following steps: manual segmentation of bones (T1W sequence), automated detection of reference signal region, sacroiliac joint central lines and ROIs, a division of ROIs into quadrants, automated detection of inflammatory changes (STIR sequence). As a gold standard, two sets of manual lesion delineations were created. Two approaches to the performance assessment of lesion detection were considered: pixel-wise (detections compared pixel by pixel) and quadrant-wise (quadrant to quadrant). Statistical analysis was performed using Spearman's correlation coefficient. Correlation coefficient obtained for pixel-wise comparison of semi-automated and manual detections was 0.87 (p = 0.001), while for quadrant-wise analysis was 0.83 (p = 0.001). The correlation between two sets of manual detections was 0.91 for pixel-wise comparison (p = 0.001) and 0.88 (p = 0.001) for quadrant-wise approach. Spearman’s correlation between two manual assessments was not statistically different from the correlation between semi-automated and manual evaluations, both for pixel- (p = 0.14) and quadrant-wise (p = 0.17) analysis. Average single slice processing time: 0.64 ± 0.30 s. Our method allows for objective detection of bone marrow oedema lesions in patients with axSpA. The quantification of affected pixels and quadrants has comparable reliability to manual assessment.

In this paper it is studied how three methods of quantifying structural anisotropy are related. Mean intercept length (MIL) method has been designed for the analysis of binary images. Autocorrelation function and the covariance matrix of the gray-level intensity gradient (GST method) are approaches designed for the analysis of gray-level data. It is shown here that the autocorrelation function and the MIL methods are not related in a general case. In contrast, an analytical proof is provided to show that MIL and GST methods are strictly equivalent. The standard definition of MIL is expressed in terms of a gradient field. Next it is shown that eigenvectors of the MIL fabric tensor are also eigenvectors of the GST fabric tensor and eigenvalues of the MIL fabric tensor can be determined if the eigenvectors of the GST fabric tensor are known. It follows from the study that the application of MIL in the assessing quality of trabecular bone can be replaced in all cases by the application of the GST method, which is more general (as defined for gray-level data), easier to implement and less computationally expensive.

In this study the relationship between the mean intercept length (MIL) method - the current standard histomorphometric method of assessing structural anisotropy and an alternative method of the gray-level structure tensor (GST) is investigated. Both methods are applied to a set of 25 three-dimensional binary μCT images of trabecular bone. It is shown that there is a very strong correlation between the logarithms of the principal values of the MIL and the GST fabric tensors (Pearson's coefficient of correlation higher than 0.98) and between the logarithms of the invariants of the MIL and the GST fabric tensors (Pearson's coefficient of correlation higher than 0.999). There is also a good correlation between the degree of anisotropy calculated from the MIL and from the GST tensors (Pearson's coefficient of correlation equal to 0.90). The principal anisotropy directions of the MIL and the GST fabric tensors coincide at the 5% significance level. Additionally, the performance of both methods is tested, based on a set of artificial structures with prescribed orientations.