Satsuki Kumasaka

Skeletal pattern extraction of bone trabeculae using mathematical morphology

Studies of morphology in extracting skeletal paterns of spiculae in mammograms served as a theoretical framework. Using studies of morphology in extraction of skeletal patterns of spiculae in mammograms as a theoretical framework, we applied a morphological filter (a combination of a skeleton operation and a single structuring element) to the extraction of skeletal patterns of bone trabeculae on computed radiographic (CR) images. Then, the comparison of the original images with the extracted skeletal patterns and the relationship between the sequence number(n) of the operation and the extracted skeletal patterns was reviewed. The comparison showed that skeletal patterns of the original image were extracted as binary images when using a morphological filter, and that, when n was small (n=0 or 1), skeletal patterns of both fine and wider bone trabeculae were extracted. Furthermore, with an increase in n, skeletal patterns of fine bone were eliminated, and only wider skeletal patterns were extracted selectively. These findings suggest that development of a suitable morphological filter to extract the skeletal patterns of bone trabeculae may be an effective supplement for computer aided diagnosis of medical images (CADM) in the support of adequate diagnosis of bone diseases.

Morphologically extracted trabecular skeleton superimposed upon digital radiograph structure

Abstract: The purpose of this study was to employ a morphological filter to digital X-ray images to extract the morphology of trabecular structures in a clearly understandable visual format. This study compares the trabecular skeleton extracted by a morphological filter to the original digital radiographic image by superimposing the images. A morphological filter (a combination of a single structuring element and a skeleton operation) based on a mathematical morphology theory was used to extract the skeletal pattern of trabecular bone from a digital X-ray image of a human femoral neck. Subset images with different operation numbers (n = 1, n = 2) were obtained, and then each image was superimposed on the original digital radiographic image using the superimpose function of a workstation. The extracted trabecular skeleton pattern was fairly consistent with the trabecular structure seen on the digital image according to the opinion of seven dentomaxillofacial radiologists. In their opinion, the majority of the structural elements were reproduced on the extracted skeleton. However, accurate skeleton elements were not extracted in the region of dense trabecular structure. The morphological filter was able to extract a large portion of the bone trabecular structure as a binary skeletal pattern image from trabecular bone on digital X-ray image, but more work is needed to improve the assessment of dense trabeculae.

Assessment of bone feature parameters from lumbar trabecular skeletal patterns using mathematical morphology image processing

Abstract. In this study, bone feature parameters were obtained from digital lumbar vertebral images to determine their potential usefulness in assessing binary trabecular skeletal patterns. The system consisted of a magnification radiographic technique, computed radiography (CR), microfocused X-ray computed tomography (μ-CT), and mathematical morphology image processing. A digital CR image was produced, using a ten-times magnified lumbar vertebral cancellous bone block (1.5 × 1.5 × 1.5 cm3). The information was then subjected to mathematical morphology image processing, to extract eight binary trabecular skeletal patterns having different continuities as the operation number (n) increased. These skeletal patterns were used as test patterns and analyzed quantitatively by calculation of the skeletal pixel percentage (SKP) and star volume analysis (Vsk, volume of skeletal trabecular elements; Vsp, volume of nonskeletal elements). Then the binary skeletal pattern data were converted into the μ-CT format, using a general personal computer, and the images were quantitatively analyzed with μ-CT built-in image analysis software for microstructural indices, fractal dimension analysis, and node-strut analysis. The SKP and Vsk showed an expected decrease as n increased. Concurrently, the Vsp showed an expected increase as n increased. The skeletal thickness (Sk.Th) showed a constant value at n= 1 to 4 and n= 5 to 7, but decreased in a stepwise fashion at n= 1 and n= 5. The skeletal number (Sk.N), bone area fraction (B.Ar/T.Ar) and bone perimeter fraction (B.Pm/T.Ar), decreased as n increased. Skeletal space (Sk.Sp) and perimeter-to-area ratio (B.Pm/B.Ar) increased with increasing n. In the fractal dimension analysis, the values decreased with increasing n and showed changes similar to the image observations. In fact, the SKP, star volume analysis, Sk.N, B.Ar/T.Ar, B.Pm/T.Ar, Sk.Sp, and B.Pm/B.Ar all closely mirrored the observational analysis of the images. Linkage of the skeletal structure as functions of the following parameters could be quantitatively demonstrated. The parameters used were node (Nd) and terminus (Tm). Variations in the total strut length (TSL) and Tm could demonstrate quantitative changes in skeletal structure. These results indicate that the system consisting of a combination of a magnification radiographic technique, CR, μ-CT, and mathematical morphology image processing may be a useful tool for quantitative skeletal structure analysis and the structural assessment of lumbar vertebrae, for the assessment of skeletal structural changes. It is important to choose suitable parameters for the desired structural changes.

Relationship between bone mineral density and bone stiffness in bone fracture

ObjectivesUsing cancellous bone blocks of racehorses, the relationship between bone mineral density (BMD), which indicates bone strength, and stiffness in bone fracture occurrences was studied.MethodsTwo groups of cancellous bone blocks were prepared: a fractured group, using the first phalangeal bones of seven racehorses with sagittal fractures; and a nonfractured group, using the first phalangeal bones of nine autopsied racehorses without any fractures. By a peripheral quantitative computed tomography scan, the BMD values were shown as color images and evaluated. In addition, the BMD values obtained from the fractured and nonfractured groups were compared with the stiffness values obtained from a compression test.ResultsThe difference between the average BMD values of the fractured and nonfractured groups was easily observed on the BMD color-conversion display image. The average BMD of the fractured group (472.1 mg/cm3) was significantly higher than that of the nonfractured group (284.5 mg/cm3, P = 0.005). Moreover, the average stiffness of the fractured group (5564.5 N/cm) was significantly higher than that of the nonfractured group (3808.6 N/cm, P = 0.008).ConclusionThese results suggest that the occurrence of a fracture does not depend on the BMD or the bone stiffness value.

Comparison of extracted skeleton patterns of human femur and humerus using the line skeleton operation process and skeleton operation process

This study presents and compares trabecular skeletal patterns of human femoral and humeral bone extracted by applying mathematical morphology and morphological filters. Two morphological filters and skeleton processing operations were used to extract the skeleton images. Images obtained with a combination of the single disc-shaped structuring element (5 pixel diameter) and the skeleton operation were compared to images obtained with the line structuring element and the line skeleton operation. Seven dentomaxillofacial radiologists evaluated the resulting images. Line skeleton images (both subset and sumset images) had fewer trabecular elements extracted and higher discontinuity as the operation number increased, and there were obvious differences between the simple skeleton images and the line skeleton images. With more study these differences may be useful in detecting changes in trabecular patterns resulting from disease, stress or altered function.

Quantitative analysis of bone trabecular structural change using X-ray bone morphometric analysis: a correlation between structural change and structural parameters

ObjectivesTo develop an X-ray bone morphometric analysis method using a combination of morphological filter processing and bone histomorphometry.MethodsPhantoms for the evaluation were prepared from a 10 × 10 × 10 mm block of cancellous bone removed from a human lumbar spine and cut into five sections, each 2 mm thick. Also prepared were 1.5-mm-thick phantoms of either cortical bone or soft tissue (X-ray tube side 10 mm, imaging plate side 5 mm). Standardized digital radiography was performed at each of five thicknesses of the cancellous bone without changing the total thickness. This was accomplished by serially replacing the 2-mm sections of the bone blocks with the soft-tissue phantoms. After morphological filter processing of the digital images, the skeletal patterns of the bone trabeculae were extracted and quantitatively assessed by bone histomorphometric analysis. The accuracy of the correlations between the trabecular structures prepared by changing the cancellous bone thickness and several structural parameters were evaluated.ResultsEach parameter obtained by the bone histomorphometric analysis showed a correlation coefficient ® of more than 0.89 between the cancellous bone thickness and the analytic value. Star volume analysis demonstrated a close correlation (r = 0.97) between the skeletal star volume and the bone thickness. Node-strut analysis indicated that the correlation coefficients of all of the parameters were larger than 0.93.ConclusionX-ray bone morphometric analysis that combines morphological filter processing and bone histomorphometry is clinically applicable as a computer-aided diagnosis for the evaluation of trabecular structures.

Radiological Morphometric Analysis of the Mandibular Bone Structure after Ovariectomy in Mature Cynomolgus Monkeys

Abstract [Purpose] The effects of experimental osteoporosis on the trabecular bone structure of the mandible in cynomolgus monkeys were examined by radiological bone morphometric analysis. [Methods] Ovariectomy (OVX) was performed on twelve 10-year-old female cynomolgus monkeys, which were fed a controlled diet for 2 years. Twelve monkeys in a sham control group were fed under the same conditions. Using a microfocus tube and computed radiography, the removed mandibular bone samples were imaged by standardized magnification radiography, and two-dimensional digital imaging data were obtained. The structural parameters, such as skeletal area, perimeter, number, complexity, continuity and anisotropy, were measured using radiological bone morphometric analysis. The bone mineral density (BMD) was measured by dual energy X-ray absorptiometry. The width of the cortical bone was measured using magnified radiographic images. [Results] There were no significant differences between the OVX and the sham control groups in the skeletal structure indicated by the skeletal volume, number, width, perimeter, complexity, continuity, separation and spacing. However, there were significant differences between the two groups in the BMD of the mandibular body, cortical bone width, anisotropy and some parameters of the skeletal continuity. Among these parameters, the difference in the thinning of the cortical bone was most significant. [Conclusions] Using two-dimensional digital radiographic image data, this study suggests that the cortical bone width is more useful than the trabecular bone structure as the morphologic parameter for diagnosis of osteoporosis in the mandibular body.

Evaluating two-dimensional skeletal structure parameters using radiological bone morphometric analysis

ObjectivesTo investigate the reliability of two-dimensional (2D) skeletal structure parameters obtained using radiological bone morphometric analysis.MethodsThe 2D skeletal parameters in the regions of interest (ROIs) were measured on computed radiography (CR) images of first phalanges from racehorses, using radiological bone morphometric analysis. Cancellous bone blocks were made from the phalanges in the same position as the ROI determined on CR images. Three-dimensional (3D) trabecular parameters were measured using microcomputed tomography (µCT). The correlations between the 2D skeletal parameters and 3D trabecular parameters were evaluated in relation to the measured bone strength.ResultsThe following 2D skeletal structure parameters were correlated with bone strength (r = 0.61 − 0.69): skeletal perimeter (Sk.Pm), skeletal number (Sk.N), skeletal separation (Sk.Sp), skeletal spacing (Sk.Spac), fractal dimension (FD), and skeletal pattern factor (SkPf). The 3D trabecular structure parameters were closely correlated with bone strength (r = 0.74 − 0.86). The 2D skeletal parameters Sk.N, Sk.Pm, FD, SkPf, and Sk.Spac were correlated with the 3D trabecular parameters (r = 0.61 − 0.70).ConclusionsThe 2D skeletal parameters obtained using radiological bone morphometric analysis may be useful indicators of trabecular strength.

Application of node-strut analysis to skeletal patterns on digital radiographic images

In this study, we applied node-strut analysis to digital images of known quantitative structures to determine its potential usefulness in assessing binary skeletal patterns extracted by morphological filtering of digital radiographic image data. The procedures we used included computed radiography (CR), node-strut analysis, and a mathematically based morphologic filtration. Six metallic wire lattice structures were so assembled as to have varying node (Nd) and terminus (Tm) points, and the lattice structures were used as test patterns whose structure was sequentially modified by reducing the wires in 6 steps. Digital radiographic images of the test patterns were produced using CR, and then 12 binary skeletal patterns having different numbers of Nd and Tm and varied strut length were extracted by the morphological filter. The binary skeletal pattern data were processed into the image data of 1-pixel thickness by using the thinning operation for node-strut analysis and then quantitatively assessed by node-strut analysis.In each of the skeletal test patterns, the theoretical values and the values produced by the node-strut analysis were well correlated. The values of the analysis were also in agreement with our visual observation.These results prove that the differences of connectivity of binary skeletal patterns extracted by morphological filtration may be assessed through the use of node-strut analysis of the numbers of Nd and Tm and the length of the strut.

Relationship between bone mineral density and 2D and 3D structural parameters of bone trabeculae

ObjectivesWe analyzed the two- (2D) and three-dimensional (3D) structure of trabecular bone using bone mineral density (BMD) values as the standard. Correlations between the 2D and 3D structural parameters and BMD in the same samples of trabecular bone were determined to assess the reliability of 2D radiological morphometric analysis of bone as a method for evaluating BMD.MethodsThe first phalanxes of racehorses were used. A region of interest (ROI) was specified on computed radiography images of the phalanx. The 2D structural parameters for the ROI were determined by morphometric analysis. Then, a block of trabecular bone corresponding to the ROI was excised from the phalanx, and the 3D structural parameters of this bone sample were determined using microfocus X-ray computed tomography. The BMD of each bone sample was measured using peripheral quantitative computed tomography.ResultsMost, but not all, of the 3D parameters were correlated with BMD (r = 0.82–0.98), and several 2D parameters (skeletal perimeter, skeletal number, skeletal separation, skeletal spacing, fractal dimension, and skeletal pattern factor) were relatively closely correlated with BMD (r = 0.57–0.65).ConclusionsThese results suggest that the 2D morphometric analysis of bone is a useful, noninvasive method for assessing bone strength.