Bone measurement by enhanced contrast image analysis Ovariectomized and intact Macaca fascicularis as a model for human postmenopausal osteoporosis.код для вставкиСкачать
AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 67:99-103 (1985) Bone Measurement by Enhanced Contrast Image Analysis: Ovariectomized and Intact Macaca fascicularis as a Model for Human Postmenopausal Osteoporosis E.A. BOWLES, D.S. WEAVER, F.W. TELEWSKI, A.H. WAKEFIELD, M.J. JAFFE, AND L.C. MILLER Departments of Anthropology @,A.B.,D.S. W.,L C.M.),Biology (A.H. W , M.JJ.),and Comparadiue Medicine (D.S. U?,L C.M.), Wake Forest Uniuersity, Winston-Salem, North Carolina 27109, Laboratory of TreeRing Research, University of Arizona, Tucson,Arizona 85721 (Z? W.Z) KEY WORDS Image analysis, Radiographic density, Postmenopausal osteoporosis, Macaca fascicularis ABSTRACT This paper presents a n image enhancement and analysis system (DARWIN) based on a n inexpensive microcomputer and applies the system to two bone morphometry problems relevant to postmenopausal osteoporosis. Using ovariectomized and intact female Macaca fascicularis as a model, we examined the radiodensity of the sixth lumbar vertebra and the cross-section area of the right femur. Significantly lower bone density was observed in the vertebral segments of the ovariectomized animals. No significant differences were observed in comparisons of the femoral cross sections. The reduction in radiographic density of the ovariectomized animals' vertebrae is similar to that observed in postmenopausal women, supporting the use of female cynomolgus macaques as models of bone loss in postmenopausal osteoporosis. Osteoporosis in human beings is character- (Cameron and Sorenson, 1963; Gordan and ized by the accelerated loss of bone, without Vaughan, 1976). Total mineral content in the changes in chemical composition, ultimately axial skeleton without distinction between often resulting in femoral neck, distal radius, cortical and trabecular bone can be meaand vertebral wedge fractures (Albanese, sured using dual photon absorptiometry and 1977). Osteoporosis is a serious health prob- neutron activation analysis (Cann et al., lem with considerable social and medical im- 1980). The absorptiometry methods inform plications; the national cost associated with on the density of bone but provide no details osteoporosis was estimated to be a t least 3.8 of the structure of the bone. Quantitative billion dollars annually in the United States computed tomography can distinguish be(NIH,1984).Osteoporosis is most pronounced tween trabecular and cortical bone mineral among post menopausal wamen and is asso- contents, which change at different rates in ciated with the hormonal changes accompa- osteoporosis, and provides a better estimate nying the cessation of ovarian function of the extent of vertebral osteoporosis (Cann et al., 1980). (Raisz, 1982). To evaluate the histological effects of menEarly detection of the osteoporotic condition without painful intrusive methods is dif- opause in bone, intrusive bone biopsies must ficult and fraught with error. Standard be performed. Standard morphometrics on radiographs can be used to measure relative various bone types and bone microstructure densities by radiogrammetry or micro- can be performed with manual or automatic densitometry (Garn, 1970). However, bone digitizers and eyepiece micrometers (Teitelpositioning, tissue interference, film quality, baum et al., 1976; Dempster et al., 1983; and film exposure present uncontrollable Smith and Jee, 1983). Microradiographic variables in the radiographic methods of analysis. Photon absorptiometry measuring bone mineral content by absorption of radiaReceived December 16, 1983; revised October 26, 1984; action from the isotope 1-125 is a useful method cepted October 30,1984. 0 1985 ALAN R. LISS, INC. 100 E.A. BOWLES ET AL. Fig. 1. Digital AnaIyzer of Resolvable Whole-picturesby Image Numeration (DARWIN). analysis can detect mineral content in individual bone microstructures (Stout and Simmons, 1979). This paper will use a new microcomputerbased automated image analyzer, of local design, to examine the effects of ovariectomy on radiographic bone density in a sample of Macaca fmcicularis. MATERIALS AND METHODS Thirty-eight adult female cynomolgus macaques (Macaca fmcicularis Medway) were skeletonized by maceration. Seventeen of the animals had been ovariectomized 22 months prior to death. Housing medical and diet conditions were identical €or both the intact and the ovariectomized animals. Mean body weights of the two groups of animals (intact and ovariectomized) were not significantly different. Hormonal assays showed significantly lower serum estradiol and progesterone levels at necropsy in the ovariectomized females (Weaver and Brodish, 1983).This paper will address one aspect of the project, which was to assess the utility of Macaca fmcicularis as a n animal model for human postmenopausal osteoporosis by examining relative bone density differences in the sixth lumbar vertebrae and in bone surface area differences in femoral cross sections. To accomplish the assessment, the authors used the Digital Analyzer of Resolvable Whole-pictures by Image Numeration (DARWIN), a n image enhancement and analysis system (see Fig. 1). The analog signal from a Dage model VC-65-S vidicon camera is converted to a digital signal by a Colorado Video model 270A digitizer which is in communication with a n Apple 11+ microcomputer through a n interface board of local design (Telewski et al., 1983). The digitizer has a resolution of 512 lines horizontally by 480 lines vertically to produce 245,760 matrix coordinates (pixels). This matrix can be accommodated by the microcomputer and the camera, and has 256 gray-scale levels, a black level adjustment, and amplitude adjustment. The video image can be enhanced to optimize contrast using the externalized video gain and target and gamma functions of the camera. The DARWIN system can analyze microscopic and macroscopic objects, photographs, and negatives. After the external controls are adjusted, the microcomputer is calibrated to a standard length, and the scan field “window” is preset. During the scan, the digitized signal is read by the microcomputer according to the program in use. The computer compares each pixel of interest to the gray scale and saves its respective grayscale value. The duration of a scan depends on the size of the scan window. The longest a scan of a n entire screen would take is 17 seconds. To measure relative radiographic densities of sixth lumbar vertebrae, the DENSITY SCAN program was used. DENSITY SCAN is a program of local design which allows DARWIN to function as a scanning densitometer. Since gray-scale values were de- IMAGE ANALYSIS OF OSTEOPOROTIC MACAQUE BONE sired, contrasts were not maximized; the camera was adjusted to measure over the entire range of gray-scale values. DENSITY SCAN does not measure absolute density. The vertical scale of the density profile (Fig. 2) is the percent gray scale or the percent transmittance of light through the measured object. In this study, a 100%F a y scale value corresponded to complete opacity, or highest bone density. The sixth lumbar vertebra from each animal was bisected transversely, and the superior and inferior vertebral halves were radiographed on a single large (35 x 28 cm) radiographic plate at 45 kV, 0.01 mas, 1 m focal distance. To reduce orientation error, all vertebral halves were positioned with the cut surface against the radiographic cassette. The radiograph was placed on a light table under the video camera. The scan window was set to cover a 2-mm-widefield along the midline of the vertebral image, from the dorsal to the ventral borders. A razor blade was placed across the spinous process of the vertebra being measured to provide a black-level standard (zero % on the gray scale).The video amplitude was adjusted so the gray scale value of the vertebral foramen equalled the zero value of the razor blade. We averaged three scans, for a total scan time for each vertebra of 6 seconds. DENSITY SCAN then displayed the density profile (Fig. 21, printed the profile with numbered peaks, the peak height of each numbered peak of interest, the area under any designated segment of the density profile, and stored the data. Relative density differences between the intact and ovariectomized groups of animals were evaluated from the density profiles using various measures (Fig. 2). The dorsal and ventral peak heights represent the regions of greatest bone density through the vertebral borders. The lowest trough height represents the region of least bone density through the vertebral body. These measures, standardized by dividing each by the thickness of the cut vertebral body at its midpoint, were compared between the two groups. Cross sections of midshaft right femora (0.5 mm thickness) were then scanned with DARWIN using the TOPOGRAPHER I1 program to determine bone surface area. TOPOGRAPHER 11 is a program of local design which recognizes and selects topographical features of interest on the basis of preset gray-scale limits. The image must be enhanced to provide sufficient contrast between the features 101 DENSTY TRACE AND VERTEBRA HRRK)R VIEW OF Fig. 2. Vertebral scan. of interest and the background. The femoral sections, mounted on glass slides, were placed on the light table under the video camera. The scan window was selected,the object and background levels set, and the instrument calibrated to a millimeter scale. After the scan, which took approximately 7 seconds for each cross section, a high-resolution reconstruction of the cross section was displayed and compared to the actual image. Dirt or other extraneous objects were eliminated before the scan data were stored. TOPOGRAPHER II displays, stores, and prints out the area, perimeter, and gray scale value of each topographical feature of interest, and the percent of the total scan area occupied by the objects of interest. The DARWIN system has been tested for precision in density and area measurements. An EIA Logarithmic Reflectance Chart (Hale Color Consultants), consisting of nine different gray-scale chips of known reflectance, was scanned, producing a density trace with “steps” corresponding to different chip reflectances. The average height of each density trace step was compared to the % reflectance of the corresponding chip to determine the accuracy of the density measures. To test the 102 E.A. BOWLES ET AL. TABLE 1. Bensitometric measurements an sixth lumbar vertebral halves' Superior half Dorsal peak height Lowesttrough height Ventral peak height Inferior half Dorsal peak height Lowest trough height Ventral peak height (n = 17) ovx Intact (n = 21) Percent difference2 4.31 (0.53) 1.31 (0.23) 4.78 (0.37) 4.83 (0.33) 1.61 (0.18) 6.26 (0.52) 10.77 18.63 23.64 > .10 > .10 0.25 > P > ,010 4.72 (0.37) 1.82 (0.24) 6.33 (0.47) 6.15 (0.45) 2.42 (0.21) 7.95 (0.44) 23.25 24.79 20.38 ,010 > P > .005 P .05 > P > ,025 ,010 > P ,005 'Mean (S.E.).OVX, ovariectomized. %tact density - OVX density x 100. intact density TABLE 2. Femoral cross-section surface area' Cross section (mm') Body weight (kg) Cross sectionhody weight ovx Intact (n = 21) (n = 17) P 38.94 (1.10) 3.07 (0.11) 12.86 (0.37) 37.24 (1.0) 2.97 (0.11) 12.69 (0.37) P > .10 P > .10 P > .10 'Mean (S.E.). precision of the density measures, density scans of seven radiographs of Plexiglas density wedge were performed. The mean step height generated by each change in density was correlated to the natural log of the actual height of the wedge step as measured with dial calipers. To evaluate the precision of the area measurements, a 100-mm2square and a metal disk of known area (283.53 mm2) were each scanned 20 times by the TOPOGRAPHER I1 program, and the results compared to the actual areas. RESULTS The DARWIN system of image analysis provides a very high level of precision and accuracy in density and area measurements. After 20 scans, the average DARWIN area of a 100-mm2square was found to be 99.91 mm2 & 0.11,with the average percent deviation 0.42 f 0.07. The average area of the metal disk, after 20 trials, was 282.96 mm2 0.060 with an average percent deviation of 0.22 f 0.010. Using the EIA Logarithmic Reflectance Chart, the correlation between the gray-scale chip % reflectance and the natural log of the mean step height of the chip on the density trace was highly significant (r = .992, P < .005). In the density precision analysis, the correlation of the plexiglass wedge step height to the density trace step height was .9899 (P < .005). The analysis of bone density of the sixth lumbar vertebrae of the intact and ovariectomized animals showed differences in the reIative amounts of bone present. The relative amount of bone present is represented by the gray-scale density reading divided by the animal's weight (kg). In all cases, the ovariectomized female macaques had lower mean densities for the three measurements in both the superior and inferior vertebral halves (Table 1). The superior half-ventral peak heights in the ovariectomized animals were significantly lower than those of the intact group. All three inferior half measurements were significantly lower in the ovariectomized female macaques than in the intact females. The ovariectomized macaques had lower mean femoral cross-section area than the intact animals, but the difference was not statistically significant (Table 2). DISCUSSION The DARWIN system has proven useful in evaluating the utility of Macaca fascicularis as a potential animal model for human postmenopausal osteoporosis. The DARWIN system can produce highly accurate density and area measurements. The high correlation between the density trace step height of the IMAGE ANALYSIS OF OSTEOPOROTIC MACAQUE BONE density wedge on seven different radiographs and the actual step heights shows that differences in radographic quality have little effect on density measures. Although the between-groups difference of the femoral cross-section surface area was not statistically significant, the direction of the difference, with the ovariedomized females having slightly less cortical bone, was interesting, especially in light of the identical conditions under which the two groups were housed. A longer postovariectomy experimentation duration in future studies might result in greater differences in the bone cross-section surface area. The density data for the sixth lumbar vertebrae show a significantly lower amount of bone in the ovariectomized animals. The percentage density difference between the two groups ranges from 10.77% to 24.79% and is consistent with the pattern of lumbar vertebral mineral loss in ovariectomized women (Cann et al., 1980). The data on differential bone density show less bone present in the ventral region of the superior segment of the vertebrae. This pattern of bone density would encourage eventual wedge fractures, of the kind occasionally seen in human postmenopausal osteoporotics (Raisz, 1982) CONCLUSIONS The use of a n image analyzer to examine the degree of bone density difference in intact and ovariectomized female Macaca fas cicularis indicates that the cynomolgus macaque may prove a useful animal model for postmenopausal osteoporosis. The cynomolgus macaque exhibits regular, monthly ovarian cycles that are similar in length to human females (Napier and Napier, 1967) and shows depression of estrogen levels following ovariectomy (Weaver and Brodish, 19831, a s do human females. In this study, the lumbar vertebrae of the ovariectomized monkeys showed lower radiographic density than those of the intact animals. The degree of density difference agrees well with vertebral density measurements taken in studies on ovariectomized human females. As in human females, substantial bone loss has occurred in the vertebral column of the 103 ovariectomized macaques with no appreciable effect in femoral midshaft cortical bone. ACKNOWLEDGMENTS This research was supported by NIH grant RROO919-10 and NHLBI grant HL 14164 to the Department of Comparative Medicine and by National Science Foundation grant PCM-8003689 and National Aeronautics and Space Administration grant NAGW-96 to M.J.J. We are greatly indebted to Dr. R. Biro for his help in performing the precision measurements on the DARWIN system. LITERATURE CITED Albanese, AA (1977)Bone loss: Cause, detection, and therapy. In: Current Topics in Nutrition and Disease, Vol 1. New York: A.R. Liss, pp. 75-100. Cameron, JR,and Sorenson, J (1963)Measurement of bone mineral in uiuo: An improved method. Science 142:122-130. Cann, CE, Genant, HK, Ettinger, B, and Gordan, GS (1980)Spinal mineral loss in oophorectomized women. J.A.M.A. 18:2056-2059. Dempster, DW, Arlot, MA, and Meunier, PJ (1983)Mean wall thickness and formation periods of trabecular bone packets in corticosteroid-induced osteoporosis. Calcif. Tissue Int. 35:410-417. Garn, SM (1970)The Early Gain and Later Loss of Cortical Bone. Springfield C.C. Thomas. Gordan, GS, and Vaughan, C (1976)Clinical Management of the Osteoporoses. Acton: Publishing Sciences Group. Napier, JR,and Napier, PH (1967)A Handbook of Living Primates. New York Academic Press. National Institutes of Health (1984)Osteoporosis. Consensus Development Conference Statement, Vol. 5, No. 3.Bethesda: NIH. Raisz, LG (1982)Osteoporosis. J. Am. Geriatr. SOC. 30:127-138. Smith, JM, and Jee, WSS (1983)Automated skeletal histomorphometry. In RR Recker (ed): Bone Histomorphometry: Techniques and Interpretation. Boca Raton: CRC Press. Stout, S, and Simmons, DJ (1979)Use of histology in ancient bone research. Yrbk. Phys. Anthropol. 22:228- 249. Teitelbaum, SL, Rosenberg, EM, Richardson, CA, and Avioli, LV (1976)Histological studies of bone from normocalcemic postmenopausal osteoporotic patients with increased circulating parathyroid hormone. J. Clin. Endocrinol. Metab. 42537-543. Telewski, FW, Wakefield, AH, and Jaffe, MJ (1983)Computer-assisted image analysis of tissues of ethreltreatedPinus taeduseedlings.Plan. Physiol.72:177-181. Weaver, DS, and Brodish, DL (1983)A model for postmenopausal osteoporosis using Mucucu fasciculuris. Am. J. Phys. Anthropol. 60:269 (Abstract).