- Email: [email protected]
Quantitative assessment of abdominal aortic artherosclerosis observed in CT scans
Computerized Medical Imaging and Graphics, Vol. 21, No. 3, pp. 185 193, 1997 © 1997 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0895~111/97 $17.00 + .00
Pergamon
PII: S0895-6111(97)00003-7
QUANTITATIVE ASSESSMENT OF ABDOMINAL AORTIC ARTHEROSCLEROSIS OBSERVED IN CT SCANS
Geoffrey Dougherty* Faculty of Allied Health Sciences, Kuwait University, P.O. Box 31470, Sulaibikhat 90805, Kuwait (Received 9 July 1996; revised 30 January 1997) Abstract--Calcification of the aorta is characteristic of artherosclerotic disease. A novel numerical index for quantitating the extent of calcification observed in computed tomography (CT) scans has been developed. The index weights the fractional area of the aorta showing calcifications with the average calcium density in the plaques. It obviates the need for manual tracing of the calcifications, and avoids the arbitrary nature and lack of reproducibility of the "calcium score" system currently used in assessing CT scans. Continuously different levels of calcification were distinguished in the abdominal aortas of a group of Kuwaiti females (n = 20), with a linear correlation coefficient between index value and age of 0.914 (P < 0.0001) over three decades. The index is generally applicable to all arterial calcifications, and could be used to monitor the effect of therapeutic regimes. It could easily be modified for use with 3D reconstructions obtained from CT helical scanning or MRI. © 1997 Elsevier Science Ltd.
Key Words: Artherosclerosis, Abdominal aorta, Calcifications, Computed tomography
sequences have been used to identify and characterize different tissue types in the arterial wall, an essential prerequisite for staging the disease (4, 5). MRI may well become the modality of choice, but its use is currently limited by financial considerations. Ultrasonography is non-invasive and has long been used to characterize blood flow in real time. Although spectral analysis of the flow is essential to determine stenosed states and verify occlusion (6, 7), measurements of the intima-media thickness are useful for detecting early artherosclerotic changes of the arterial wall. Calcification of the arterial wall results in typical acoustic shadowing (8) and morphometric descriptors of the plaque surface may permit a more rigorous analysis of lesion morphology. More recently, 3D-imaging by ultrasound has proved useful in demonstrating the position of selected carotid plaques and the degree of lumen narrowing (9). Functional monitoring of artherosclerosis using radioisotope techniques is still in its infancy, and the limited reaction time of cellular blood constituents with the vascular surface at the lesion site is hampering development (10). Currently, clinical evaluation of aortic artherosclerosis is most commonly performed by visual analysis of aortic calcification on conventional radiographs or, more recently, on CT scans. Although calcifications can be detected easily, precise evaluation of the degree of calcification, such as would be required in order to measure small changes with increasing age, is rarely attempted. Instead, either an arbitrary grading
INTRODUCTION
Cardiovascular disease is the major cause of disability and mortality in developed countries (1), and early diagnosis and characterization of the underlying artherosclerosis is an essential element in limiting future mortality and/or morbidity. Artherosclerosis is initiated primarily by artheroma formation in the inner wall of the artery due to cholesterol deposition, followed by the deposition of fibrous protein and cellular material from the blood. The fibrous plaques may become ulcerated due to degenerative changes of the elastin and/or collagen fibres in the plaque, and eventually the lesion becomes calcified. The resulting occlusion of the artery is one of the most frequently occurring abnormalities of the cardiovascular system, and can result in vascular ischemia, tissue infarction and thrombosis. There are relatively few tests which are sensitive enough to identify individuals most at risk from symptomatic heart disease. Testing of the heart under stress using ECG, echocardiography or nuclear imaging is the most common procedure for followup of high blood cholesterol or other risk factors. Xray angiography clearly shows the presence of plaque, but is an invasive and expensive procedure only justified after severe cardiovascular problems have already developed (2, 3). Various MRI pulse *Fax: (965) 4830937; E-mail: [email protected] 185
186
Computerized MedicalImagingand Graphics
system (11) or a "calcium score" based on the size and density of calcium deposits (12, 13) is routinely used. This paper proposes a new quantitative index for evaluating the severity of artherosclerotic calcification observed in CT scans, which takes into account the fractional area of the calcifications and the concentration of calcium within them. PATIENTS AND METHODS Patients
Twenty patients, aged 42-72 yr, were retrospectively identified as having abdominal aortic calcification in CT scans taken at the level of the L3 and/or L4 lumbar vertebrae. The patients were all female and Kuwaiti. They formed a rather homogeneous group in terms of weight (75-90 kg), diet (excessively carbohydrate-based) and exercise patterns (little, if any). They had been originally scanned in order to measure their trabecular bone mineral density for suspected osteoporosis: aortic calcification was noticed incidentally.
May-June/1997,Volume21, Number 3 calcification, CTi, and the mean CT number of the artery (measured in a region of interest interior to the calcifications, and presumed to be typical of the arterial lumen), viz: Wi = (CTi - C T )
(la)
since increasingly serious calcifications will be reflected in an increasingly large weight. The measurement process could be simplified by obtaining a global calcification weight for the whole arterial image, by obtaining the average value, CT, of a "shifted CT number" histogram. The "shifted" histogram is obtained by taking the histogram of a region of interest (ROI) circumscribing all the calcifications, but neglecting values below a threshold, which is the highest value indicating flowing blood and normal tissue. The appropriate threshold value is the largest CT number appearing in a histogram of a ROI of the artery interior to the calcifications. Thus, the "shifted" histogram depicts only those pixels corresponding to calcifications. The artherosclerosis index can then be written in the form
Imaging parameters
Images were obtained using a General Electric Sytec 3000 CT scanner using a protocol appropriate to bone mineral density measurement (viz. 100 kVp and 100-130 mA, a scan time of 3 s, a slice thickness of 5 mm and a scan field of view (FOV) of 35 cm). Patients were positioned with the lumbar spine over a standard mineral-equivalent phantom (QCT phantom, Image Analysis Inc., Irvine, CA) comprising three different densities (0, 75 and 150 mg cm -3) of calcium hydroxyapatite. All scans were obtained parallel to the vertebral endplates through appropriate angulation of the gantry, and were centered at the midplane of the vertebral body (L3 or L4). The digital images were 512x512 pixels in size with 12-bit quantization.
AI= CT x
AI-
~(a,.wi) A '
(2)
which can be taken as a dimensionless constant (although, strictly, it has units of HU). Hounsfield units are a relative scale, which depend on the effective energy of the X-ray beam. If severity indices are to be compared across different CT machines, and different exposure factors, it would be preferable to convert the raw mean CT number, CT, into a calcium equivalent value (C---~,in mg cm-3). This can be accomplished using a calcium hydroxyapatite phantom, since X-ray diffraction (14) and infrared spectroscopy studies (15) have identified calcium deposits in the aorta to be of this form. With reference to calcium, the proposed artherosclerosis index takes on its final form
Artherosclerosis index
We propose an index, the artherosclerosis index (A/), for evaluating the severity of arterial calcification which weights the fractional area calcified by the concentration of calcium within each deposit. Thus
A
--
A I = Ca x
EA;
A
(3)
where C---ais the mean calcium (in mg cm-3) in the calcifications.
(1)
where A; are the areas of each calcification, A is the cross-sectional area of the artery, and W~ are the weighting factors. The weights, Wi, could be taken as the differences between the mean CT number of each
Data analysis
All measurements of the abdominal aorta were made on magnified (x 8) images to minimize monitor and operator errors. The window width (WW) and window level (WL) for the display were initially set at 250 and 50, respectively, for all measurements,
Quantitative assessment of artherosclerosis • G. DOUGHERTY although W L was sometimes increased later to produce sufficiently bright hard-copy images, For each image, two circular ROIs were selected--the outer one around the outer margin of the image of the aorta, and the inner one interior to the calcifications (and presumably approximating the arterial lumen). Histograms of both ROIs were obtained using a standard scanner function (Fig. l(a) and l(b)), and the maximum value of CT
187
number appearing in the inner ROI noted. (In this study, these maximum values ranged from 80 to 155.) The histogram of the outer ROI was then displayed, for CT numbers larger than the maximum obtained for the inner ROI (Fig. l(c)). This histogram now represents the pixels in the aortic image corresponding to calcifications, and the average value of this histogram represents the average CT number of these calcifications. This
Fig. 1. Histograms of the CT numbers for ROI's (a) interior and (b) exterior to the calcifications: (c) shows the "shifted" histogram of the exterior ROI (viz. the whole aorta) for CT numbers corresponding to calcifications.
188
Computerized Medical Imaging and Graphics
average CT value was converted into a corresponding calcium-equivalent value (in mg cm -3) using the known values of the phantom, and combined with the fractional area of the aorta occupied by calcifications (using standard ROI functions) to give the atherosclerosis index. These circular ROIs do not correspond to the inner and outer margins of the arterial wall (which are not perfectly circular), although they may approximate them. Their purpose is rather to contain all the calcifications between the two ROIs, with the inner ROI providing a threshold value below which pixels represent flowing blood and normal tissue and above which pixels represent calcifications. The apparent size of the calcifications changes with the settings for W W and, particularly, WL. Changes in these settings can confound attempts to manually trace the outlines of the calcifications with any degree of accuracy. In order to avoid this variability and the difficulty of manually tracing
May-June/1997,Volume 21, Number 3
with a trackerball, we used the density masking function to highlight the calcifications. The maximum CT value obtained previously for the circular inner ROI is the appropriate minimum value for the density masking function (and the maximum can be set at the maximum value for the scanner--say, 3000). The calcifications are then highlighted automatically, and their areas displayed (Fig. 2).
RESULTS Table 1 lists the relative size and calcium concentrations for calcifications in three of the patients studied, and the resulting values of the artherosclerosis index. All the measurements in the table are the mean of five independent estimates, obtained by different operators using the standard protocol described earlier and calculated using equation (3). They are reported as mean 4-1 SD.
Fig. 2. Calcifications are highlighted automatically using the density masking function.
Quantitative assessment of artherosclerosis * G. DOUGHERTY Several representative scans, showing a range of conditions from mild to severe calcification, are shown in Fig. 3 together with their calculated atherosclerosis indices. Values of the artherosclerosis index range from a b o u t 1 in mild cases to more than 50 in severe cases. The largest values occur with large, dense plaques. Values of the artherosclerosis index measured at the mid-L3 and mid-L4 levels are plotted against
189
patient age in Fig. 4. The overlaid lines are the linear regression line for both L3 and L4 indices together (i.e. the mean index) and the 95% confidence lines. For most (80%) patients, the value of the index at the mid-L4 level is the larger. The severity of the disease, as characterized by the value of the artherosclerosis index measured at either the L3 or L4 vertebral level, shows a significant correlation with age amongst this group. The results of linear regression (at P < 0.0001)
Fig. 3. A range of calcification conditions from mild to severe. The corresponding artherosclerotic indices were (a) 1.56+0.11 (b) 8.874-0.62 (c) 45.15-1-3.16 (units are mg cm-3).
190
Computerized Medical Imaging and Graphics
May-June/1997, Volume 21, Number 3
60 50
'tO p,
Q
0
L3 index
El
L4 index
0
fl
40
30
"6 0
20
n 10
0
40
45
~ 50
55
60
65
70
75
Age (years) Fig. 4. Artherosclerosis index, at mid-L3 (O) and mid-L4 (I-q) levels, plotted against patient age. The regression line (with 95% confidence bands) for all the data points is shown (y= -73.82 + 1.65"x: correlation coefficient (r) = 0.914 (p < 0.0001)).
for the artherosclerosis index measured at the mid-L3 level, the mid-L4 level, and the mean of these two indices with age are shown in Table 2. The F-statistic is the ratio of the mean square of the dependent variable to the mean square of the residuals. The size of the Fstatistic (and its probability value) is a guide to how important the independent variable is in explaining the behaviour of the dependent variable: a low probability value indicates that it is unlikely that such a large Fstatistic would have happened by chance. The intercept on the age axis indicates that calcifications are not likely below the age of 45 yr in patients such as these. DISCUSSION A useful indicator of the severity of artherosclerosis is the consequent arterial stenosis, defined as the percentage reduction in the stenosed lumen relative to the normal lumen. %stenosis = (AnAs As) , where An=cross-sectional area of normal arterial lumen and As=cross-sectional area of stenosed arterial lumen. In conventional aortography, the external diameter of the artery can be measured in a stenosed
region and in a normal region. It is assumed that lumen narrowing will follow the narrowing of the external wall (and therein lies an important limitation to the technique). Clinically significant disease is generally considered to correspond to a narrowing of the diameter by at least 50%, or the cross-sectional area by 75%: certainly 75% stenosis by area of the lumen has been shown to be hemodynamically significant (16-18). Computed tomography is recognized as being superior to aortography at detecting intimal calcification (19), which is a marker of artherosclerosis (20). The degree of aortic artherosclerosis has generally been appraised according to an arbitrary grading system. The condition was graded as mild, if calcification was noted at one site or if scattered focal calcifications were seen; moderate, if circumferential continuous calcifications or plaque were present at one or a few levels; and severe, if circumferential changes were present at multiple levels (11, 21). More recently a "calcium score" system has been developed and used as an indicator of coronary artery disease. The size and density of calcium deposits observed in ultrafast CT images were used to determine a semiquantitative assessment of the severity of coronary artery calcification in patients. The area of calcification (in mm 2) was multiplied by an arbitrary density score based on the peak CT
Quantitative assessment of artherosclerosis • G. DOUGHERTY
number in the identified deposits (12, 13). In the absence of a phantom to relate CT numbers to actual calcium densities, calcifications had to be defined arbitrarily as comprising groups of at least four contiguous pixels with values above 130 HU: arbitrary weightings were then assigned according to peak CT value---1 for a peak in the range 130199 HU, 2 for 200-299 HU, 3 for 300-399 HU and 4 for > 400 HU. The sum of each of these scores for all of the major epicardial vessels, notwithstanding difficulties in identifying some of them, was then taken as the CT-indicated severity for coronary artery disease. Spatial resolution (based on a 512 x 512 reconstruction matrix, with a field of view of 26 cm) was similar to conventional CT: ultrafast acquisition (~100 ms) was necessary to minimize respiratory motion. This method of assessment has a number of drawbacks, the most obvious being the reduction of the continuous variable of calcium content to an arbitrarily discretized index. A fourfold increase in mortality has been reported in a group of subjects with fluoroscopically detected coronary artery calcium (22): and a number of studies have shown a direct relationship between the extent (viz. size xdensity) of arterial calcification, albeit scored semi-quantitatively, and the severity of stenotic obstructions (11-13). Radiographically detected coronary artery calcium has been shown to predict both the pathologic findings (23, 24) and the angiographic narrowings (13, 25) associated with artherosclerotic heart disease. Calcium deposits indicate plaque formation, which in turn causes stenosis. The association persists in all age groups, but is most marked in younger patients (11). Serial studies, using the relatively low reproducibility of a calcium scoring system, have shown predictive value for future symptomatic disease and even the reduction of deposits with aggressive medical management (26). The poor reproducibility of the calcium scoring algorithm has been reported particularly for low scores which can have variabilities greater than 100% (20-27). This has important implications for screening programs, for example, for coronary disease. In order to reduce the variability in calcium scoring it was recognized that alternative algorithms were required that incorporate measurements from the entire artherosclerotic lesion rather than only the peak density, and that do not bracket wide ranges of densities into discrete groups (20). The artherosclerosis index proposed in this paper addresses both of these shortcomings, to provide a continuously variable index that is more objectively quantitative than either the calcium scoring system or visual analysis.
191
The mean calcium level in the calcifications, obtained using circular ROIs and a "shifted" histogram, can be obtained precisely and reproducibly. The method does not assume circularity of the artery or the lumen. Variability due to the selection and placement of the ROIs is small, and results in a precision of less than 1.5%. The area of the calcifications ( ~ Ai) can be determined precisely (to within +1%) using the density masking protocol, but difficulties in precisely tracking the outer margin of the aortic wall can result in errors up to -t-5% in the aortic cross-section (A). However, normalization of the calcification area by the aortic cross-section is considered important since the aortic cross-section typically increases by 20% between 42 and 72 yr for females (28). The total precision of +7% in the artherosclerosis index is superior to values obtained with calcium scoring of coronary calcifications. It allows us to distinguish many gradations of severity from mild (A/H1) to severe (AI~,50), examples of which are shown in Fig. 3. The data in Fig. 4 indicate that the proposed artherosclerosis index is capable of distinguishing continuously different levels of severity within a group of patients. The increased severity with age in our cohort is consistent with studies of coronary artery calcification (12, 28, 29) and of calcium in human aortas from autopsy (15). Generally larger values were observed for the index at the L4 level, closer to the bifurcation into the iliac arteries, compared to the L3 level, especially for the younger, less severely affected patients. This agrees with suggestions that intimal thickening and lipid accumulation, both precursors of calcification, develop mainly in regions of relatively low wall shear stress (30) such as the lateral walls of arterial junctions. Our limited data suggest that abdominal aortic calcifications are rare in Kuwaiti females less than 45 yr, but that severity increases monotonically and predictably with age after that. The coefficient of determination (r 2) assesses the proportion of the dependent variable's variability that is explained by the independent variable (with a maximum value of 1.0). The value of 0.818 (P < 0.001), obtained for the mean of the indices at L3 and L4 with age, indicates the strong age-dependence of abdominal aortic calcification. The correlation is much stronger than in other studies (e.g. r 2 = 0.303 for calcium mass vs age using plain X-rays of autopsied abdominal aortas from Canadian women(15)), probably due to the homogenous nature of our test cohort. The large values of the F-statistic (at low probability values) indicate that age is very important in explaining the values obtained for the artherosclerosis index. If
192
Computerized Medical Imaging and Graphics
future studies were to use our proposed artherosclerosis index, the rate of increase could be compared amongst different ethnic groups and populations. It has been suggested that in women, the prevalence and severity of aortic calcification closely parallels the degree of osteoporosis (30): it was postulated that as calcium is leached out of bones it is deposited in the vascular tree. The majority of our patient group (17 out of 20) were diagnosed as osteoporotic on the basis of their low vertebral bone density compared to age-matched controls. However, the correlation between their degree of osteoporosis (taken as the ratio of bone mineral densities for an age matched control and the specific patient) and their severity o f artherosclerosis (given by their artherosclerosis index) was poor (r=0.13). The corresponding value of the t-statistic of 0.578 indicates that the correlation is not significant at a probability level of 5% (viz. P < 0.05). The proposed index successfully meets the need for a reliable algorithm for assessing the severity of aortic calcifications and artherosclerosis. It may be useful in predicting the development of aortic aneurysms, which result from a weakening of the aortic wall that may be related to fragmentation of connective tissue at the boundaries of calcified plaques (32). The index is considered to be generally applicable to all arterial calcifications, and is not limited to the abdominal aorta, although precise measurement of the cross-sectional area (A) will be more difficult for tortuous vessels such as the iliac arteries. The spatial extent (~Ai) of epicardial coronary artery calcification is required for both calcium scoring and the proposed artherosclerosis index algorithm. It will generally be less than the aortic calcification presented here, and CT images of the heart (even with ultrafast acquisition) will be more prone to motion artefact. This will result in some additional loss of precision. Preliminary measurements on some ultrafast CT images of coronary artery calcification have indicated a precision of 1 0 - 1 5 % for less severe calcifications. The proposed algorithm offers the potential of monitoring the effect of various therapeutic regimes, and analyzing slight changes in the evolution of artherosclerotic disease in a patient with age if a series of CT scans are available. There is currently great concern over whether the increased calcification with age can be prevented or reversed. Whilst calcified collagenous plaques in the aorta may not be reversible once established, it is apparent that some calcification in smaller arteries can be controlled or reduced (33). With a sufficiently large group of patients, it may be possible to discriminate
May-June/1997,Volume 21, Number 3 between symptomatic and asymptomatic cases and use an individual's index to predict whether that patient might be at high risk of developing cardiovascular disease. The index could be easily modified to 3D reconstructions of plaques imaged by CT helical scanning or by MRI, by replacing the fractional areas by fractional volumes. Preliminary work has begun at three public hospitals in Kuwait using M R I images of calcified plaques to assess the prognostic significance of the volume algorithm. The algorithm may also prove useful in assessing the condition of other disease processes: it has been suggested that idiopathic retroperitoneal fibrosis (periaortitis) marc occur more frequently in patients whose aortic calcifications are more severe than expected for their age and sex (21). SUMMARY A novel quantitative index has been developed and used to quantitate the extent of aortic calcification observed incidentally in CT scans. The index was tested on a limited group of post menopausal females. It was able to distinguish continuously different levels of severity with a reproducibility of + 7 % , a performance superior to the discretized calcium score system widely used to date. The index showed a strong correlation between the severity of calcification and age (linear coefficient of correlation=0.914) with this group. It may be useful in predicting whether a patient might be at high risk of developing artherosclerotic disease (31).
Acknowledgement--Thanks are due to Dr D. Newman for
suggesting work in this area, to Ms Fatma Farhan and Ms Abeer Ibrahim for assisting with some of the measurements,and to Ms Char Hunt for making the original CT scans available. REFERENCES 1. Friedewald, S.T. Epidemiologyof cardiovascular disease. In: Cecil's textbook of medicine. J.B. Wyngaarden, L.H. Smith (eds), 18th edition. Philadelphia: Saunders; 1988:179-198. 2. Senkowsky,J.; Bell,W.H.; Kertein, M.D. Normal angiograms and carotid pathology. Am. Surg. 56:726-729; 1990. 3. Merickel,M.B.; Carman, C.S.; Brookeman, J.R.; Ayers, C.R. Image analysis and quantification of artherosclerosis using MRI. Comput. Med. Imag. Graph. 15:207-216; 1991. 4. Herfkens, R.J.; Higgins, C.B.; Hricak, H.; Lipton, M.J.; Crookes, L.E.; Sheldon, P.E.; Kaufman, L. Nuclear magnetic resonanceimaging ofartherosclerotic disease. Radiol. 48:161166; 1983. 5. Wesbey,G.E.; Higgins, C.B.; Hale, J.D.; Valk, P.E. Magnetic resonance applications in artherosclerotic vascular disease. Cardiovasc. Intervent. Rad. 8:342-350; 1986. 6. Ranke, C.; Creuzig,A.; Alexander, K. Duplex scanning of the peripheral arteries: correlation of the peak systolic velocity
Quantitative assessment of artherosclerosis • G. DOUGHERTY
7.
8.
9. I0. 11. 12.
13.
14. 15. 16.
17. 18. 19. 20.
21.
ratio with angiographic diameter reduction. Ultrasound Med. Biol. 18:433-440; 1992. Guo, Z.; Durand, L.-G.; Allard, L.; Cloutier, G.; Lee, H. Classification of lower limb arterial stenoses from Doppler blood flow signal analysis with time--frequency representation and pattern recognition techniques. Ultrasound Med. Biol. 20:335-346; 1994. Laurora, G.; Caesarone, M.R.; De Sanctis, M.T.; Belcaro, G. The ultrasonic assessment of carotid plaques in relation to cerebral events and damage over 3 years. Minerra Cardioangiol. 40:265--269; 1992. Delcker, A.; Dienere, H.C. Quantification of artherosclerotic plaques in carotid arteries by 3D ultrasound. Br. J. Radiol. 67:672-678; 1994. Sinzinger, H.; Virgolini, I. Nuclear medicine and artherosclerosis. Eur. J. Nucl. Med. 17:160-178; 1990. Siegel, C.L.; Ellis, J.H.; Korobkin, M.; Dunnick, N.R. CTdetected renal arterial calcification: correlation with renal artery stenosis in angiography. AJR 163:867-872; 1994. Agatston, A.S.; Janowitz, W.R.; Hildner, F.J.; Zusmer, N.R.; Viamonte, M.; Detrano, R. Quantification of coronary artery calcium using ultrafast computed tomography. J. Am. Coll. Card. 15:827-832; 1990. Breen, J.F.; Sheedy, P.F.; Schwartz, R.S.; Stanson, A.; Kaufmaun, R.; Moll, P.; Run, J. Coronary artery calcification detected with ultrafast CT as an indication of coronary artery disease. Radiol. 185:435-439; 1992. Weissman, G.; Weissman, S. X-ray diffraction studies of human aortic elastin residues. J. Clin. Invest. 39:1657-1666; 1960. Sherebrin, M.H.; Lim, H.L.; Roach, M. Jr. Calcium mass in human aortas from autopsy. Physiol. Meas. 15:281-289; 1994. Gould, K.L.; Lipscomb, K.; Hamilton, G.W. Physiologic basis for assessing critical coronary stenosis: instantaneous flow response and regional distribution during coronary hyperemia as measures of coronary flow reserve. Am. J. Cardiol. 33:8794; 1974. Azuma, T.; Fukushima, T. Flow patterns in stenotic blood vessel models. Biorheol. 13:337-355; 1976. Maun, F.C.; Herrick, J.F.; Essex, H.E.; Baldes, E.J. Effect on blood flow of decreasing the lumen of a blood vessel. Surg. 4:249-252; 1992. Limpert, J.D.; Vogelzang, R.L.; Yao, J.S.T. Computed tomography of aortoiliac artherosclerosis. J. Vasc. Surg. 5:814-819; 1987. DeVries, S.; Wolfkiel, C.; Shah, V.; Chomka, A.; Rich, S. Reproducibility of the measurement of coronary calcium with ultrafast computed tomography. Am. J. Cardiol. 75:973-975; 1995. Dixon, A.K.; Springall, R.G.; Kelsey, F.I.; Taylor, G.W. Computed tomography (CT) of abdominal aortic aneurysms: determination of longitudinal extent. Br. J. Surg. 68:47-50; 1981.
193
22. Margolis, J.R.; Chen, J.T.; Kong, Y.; Peter, R.H.; Behar, V.S.; Kisslo, J.A. The diagnostic and prognostic significance of coronary artery calcification--a report of 800 cases. Radiol. 137:609-615; 1980. 23. McCarthy, J.H.; Palmer, F.J. Incidence and significance of coronary artery calcification. Br. Heart J. 36; 1974:499-504. 24. Moore, E.H.; Greenberg, R.W.; Merrick, S.H.; Miller, S.W.; McLoud, T.C.; Shep, J.O. Coronary artery calcifications: significance of incidental detection in CT scans. Radiol. 172:711-716; 1989. 25. Hamby, R.I.; Tabrah, F.; Wisoff, B.G.; Hartgstein, M.L. Coronary artery calcification: clinical implications and angiographic correlates. Am. Heart J. 87:565-569; 1974. 26. Imagix, San Francisco, CA. Coronary artery disease and calcification, 1993. 27. Bielak, L.; Kaufmann, R.; Moll, P.; Mc Collough, C.; Schwartz, R.; Sheedy, P. Small lesions in the heart identified at electron beam CT: calcification or noise? Radiol. 192:631636; 1994. 28. Dixon, A.K.; Lawrence, J.P.; Mitchell, J.R.A. Age-related changes in the abdominal aorta shown by computed tomography. Clin. Rad. 35:33-37; 1984. 29. DeVries, S.; Wolfkiel, C.; Fusman, B.; Bakdash, H.; Ahmed, A.; Levy, P.; Chomka, A.; Kondos, G.; Zajac, E.; Rich, S. Influence of age and gender on the presence of coronary calcium detected by ultrafast computed tomography. J. Am. Coll. Card. 25:76-82; 1995. 30. Nerem, R.M. Vascular fluid mechanics, the arterial wall and artherosclerosis. J. Biomech. Eng. 114:274-282; 1992. 31. Boukhris, R.; Becker, K.L. Calcification of the aorta and osteoporosis. J. Am. Med. Ass. 219:1307-1312; 1972. 32. Sherebrin, M.H.; Bernans, H.A.; Roach, M.R. Extensibility changes of calcified soft tissue strips from human aorta. Can. J. Physiol. Pharmacol. 65:1878-1883; 1987. 33. Feldman, F. Soft tissue mineralization: Roentgen analysis. Current Prob. Diagnostic Radiol. 15:161-240; 1986.
About the Author-----~EOFFDOUGHERTY received a BSc from the
University of Manchester in 1971 and a PhD from Keele University in 1979. He has held faculty positions at universities in England, Switzerland, Malaysia and Australia, and was a Visiting Research Professor at Swarthmore College, Philadelphia. He is currently Professor of Medical Image Processing at the University of Kuwait. His research interests include medical applications of image processing and pattern recognition, multiresolutional texture measures, and the quantitation of global image quality. Professor Dougherty is a Senior Member of IEEE, a member of AAPM, and a Fellow of IEE and AIP (note: I E E - Institution of Electrical Engineers (UK); AIP--Australian Institute of Physics).
Pergamon
PII: S0895-6111(97)00003-7
QUANTITATIVE ASSESSMENT OF ABDOMINAL AORTIC ARTHEROSCLEROSIS OBSERVED IN CT SCANS
Geoffrey Dougherty* Faculty of Allied Health Sciences, Kuwait University, P.O. Box 31470, Sulaibikhat 90805, Kuwait (Received 9 July 1996; revised 30 January 1997) Abstract--Calcification of the aorta is characteristic of artherosclerotic disease. A novel numerical index for quantitating the extent of calcification observed in computed tomography (CT) scans has been developed. The index weights the fractional area of the aorta showing calcifications with the average calcium density in the plaques. It obviates the need for manual tracing of the calcifications, and avoids the arbitrary nature and lack of reproducibility of the "calcium score" system currently used in assessing CT scans. Continuously different levels of calcification were distinguished in the abdominal aortas of a group of Kuwaiti females (n = 20), with a linear correlation coefficient between index value and age of 0.914 (P < 0.0001) over three decades. The index is generally applicable to all arterial calcifications, and could be used to monitor the effect of therapeutic regimes. It could easily be modified for use with 3D reconstructions obtained from CT helical scanning or MRI. © 1997 Elsevier Science Ltd.
Key Words: Artherosclerosis, Abdominal aorta, Calcifications, Computed tomography
sequences have been used to identify and characterize different tissue types in the arterial wall, an essential prerequisite for staging the disease (4, 5). MRI may well become the modality of choice, but its use is currently limited by financial considerations. Ultrasonography is non-invasive and has long been used to characterize blood flow in real time. Although spectral analysis of the flow is essential to determine stenosed states and verify occlusion (6, 7), measurements of the intima-media thickness are useful for detecting early artherosclerotic changes of the arterial wall. Calcification of the arterial wall results in typical acoustic shadowing (8) and morphometric descriptors of the plaque surface may permit a more rigorous analysis of lesion morphology. More recently, 3D-imaging by ultrasound has proved useful in demonstrating the position of selected carotid plaques and the degree of lumen narrowing (9). Functional monitoring of artherosclerosis using radioisotope techniques is still in its infancy, and the limited reaction time of cellular blood constituents with the vascular surface at the lesion site is hampering development (10). Currently, clinical evaluation of aortic artherosclerosis is most commonly performed by visual analysis of aortic calcification on conventional radiographs or, more recently, on CT scans. Although calcifications can be detected easily, precise evaluation of the degree of calcification, such as would be required in order to measure small changes with increasing age, is rarely attempted. Instead, either an arbitrary grading
INTRODUCTION
Cardiovascular disease is the major cause of disability and mortality in developed countries (1), and early diagnosis and characterization of the underlying artherosclerosis is an essential element in limiting future mortality and/or morbidity. Artherosclerosis is initiated primarily by artheroma formation in the inner wall of the artery due to cholesterol deposition, followed by the deposition of fibrous protein and cellular material from the blood. The fibrous plaques may become ulcerated due to degenerative changes of the elastin and/or collagen fibres in the plaque, and eventually the lesion becomes calcified. The resulting occlusion of the artery is one of the most frequently occurring abnormalities of the cardiovascular system, and can result in vascular ischemia, tissue infarction and thrombosis. There are relatively few tests which are sensitive enough to identify individuals most at risk from symptomatic heart disease. Testing of the heart under stress using ECG, echocardiography or nuclear imaging is the most common procedure for followup of high blood cholesterol or other risk factors. Xray angiography clearly shows the presence of plaque, but is an invasive and expensive procedure only justified after severe cardiovascular problems have already developed (2, 3). Various MRI pulse *Fax: (965) 4830937; E-mail: [email protected] 185
186
Computerized MedicalImagingand Graphics
system (11) or a "calcium score" based on the size and density of calcium deposits (12, 13) is routinely used. This paper proposes a new quantitative index for evaluating the severity of artherosclerotic calcification observed in CT scans, which takes into account the fractional area of the calcifications and the concentration of calcium within them. PATIENTS AND METHODS Patients
Twenty patients, aged 42-72 yr, were retrospectively identified as having abdominal aortic calcification in CT scans taken at the level of the L3 and/or L4 lumbar vertebrae. The patients were all female and Kuwaiti. They formed a rather homogeneous group in terms of weight (75-90 kg), diet (excessively carbohydrate-based) and exercise patterns (little, if any). They had been originally scanned in order to measure their trabecular bone mineral density for suspected osteoporosis: aortic calcification was noticed incidentally.
May-June/1997,Volume21, Number 3 calcification, CTi, and the mean CT number of the artery (measured in a region of interest interior to the calcifications, and presumed to be typical of the arterial lumen), viz: Wi = (CTi - C T )
(la)
since increasingly serious calcifications will be reflected in an increasingly large weight. The measurement process could be simplified by obtaining a global calcification weight for the whole arterial image, by obtaining the average value, CT, of a "shifted CT number" histogram. The "shifted" histogram is obtained by taking the histogram of a region of interest (ROI) circumscribing all the calcifications, but neglecting values below a threshold, which is the highest value indicating flowing blood and normal tissue. The appropriate threshold value is the largest CT number appearing in a histogram of a ROI of the artery interior to the calcifications. Thus, the "shifted" histogram depicts only those pixels corresponding to calcifications. The artherosclerosis index can then be written in the form
Imaging parameters
Images were obtained using a General Electric Sytec 3000 CT scanner using a protocol appropriate to bone mineral density measurement (viz. 100 kVp and 100-130 mA, a scan time of 3 s, a slice thickness of 5 mm and a scan field of view (FOV) of 35 cm). Patients were positioned with the lumbar spine over a standard mineral-equivalent phantom (QCT phantom, Image Analysis Inc., Irvine, CA) comprising three different densities (0, 75 and 150 mg cm -3) of calcium hydroxyapatite. All scans were obtained parallel to the vertebral endplates through appropriate angulation of the gantry, and were centered at the midplane of the vertebral body (L3 or L4). The digital images were 512x512 pixels in size with 12-bit quantization.
AI= CT x
AI-
~(a,.wi) A '
(2)
which can be taken as a dimensionless constant (although, strictly, it has units of HU). Hounsfield units are a relative scale, which depend on the effective energy of the X-ray beam. If severity indices are to be compared across different CT machines, and different exposure factors, it would be preferable to convert the raw mean CT number, CT, into a calcium equivalent value (C---~,in mg cm-3). This can be accomplished using a calcium hydroxyapatite phantom, since X-ray diffraction (14) and infrared spectroscopy studies (15) have identified calcium deposits in the aorta to be of this form. With reference to calcium, the proposed artherosclerosis index takes on its final form
Artherosclerosis index
We propose an index, the artherosclerosis index (A/), for evaluating the severity of arterial calcification which weights the fractional area calcified by the concentration of calcium within each deposit. Thus
A
--
A I = Ca x
EA;
A
(3)
where C---ais the mean calcium (in mg cm-3) in the calcifications.
(1)
where A; are the areas of each calcification, A is the cross-sectional area of the artery, and W~ are the weighting factors. The weights, Wi, could be taken as the differences between the mean CT number of each
Data analysis
All measurements of the abdominal aorta were made on magnified (x 8) images to minimize monitor and operator errors. The window width (WW) and window level (WL) for the display were initially set at 250 and 50, respectively, for all measurements,
Quantitative assessment of artherosclerosis • G. DOUGHERTY although W L was sometimes increased later to produce sufficiently bright hard-copy images, For each image, two circular ROIs were selected--the outer one around the outer margin of the image of the aorta, and the inner one interior to the calcifications (and presumably approximating the arterial lumen). Histograms of both ROIs were obtained using a standard scanner function (Fig. l(a) and l(b)), and the maximum value of CT
187
number appearing in the inner ROI noted. (In this study, these maximum values ranged from 80 to 155.) The histogram of the outer ROI was then displayed, for CT numbers larger than the maximum obtained for the inner ROI (Fig. l(c)). This histogram now represents the pixels in the aortic image corresponding to calcifications, and the average value of this histogram represents the average CT number of these calcifications. This
Fig. 1. Histograms of the CT numbers for ROI's (a) interior and (b) exterior to the calcifications: (c) shows the "shifted" histogram of the exterior ROI (viz. the whole aorta) for CT numbers corresponding to calcifications.
188
Computerized Medical Imaging and Graphics
average CT value was converted into a corresponding calcium-equivalent value (in mg cm -3) using the known values of the phantom, and combined with the fractional area of the aorta occupied by calcifications (using standard ROI functions) to give the atherosclerosis index. These circular ROIs do not correspond to the inner and outer margins of the arterial wall (which are not perfectly circular), although they may approximate them. Their purpose is rather to contain all the calcifications between the two ROIs, with the inner ROI providing a threshold value below which pixels represent flowing blood and normal tissue and above which pixels represent calcifications. The apparent size of the calcifications changes with the settings for W W and, particularly, WL. Changes in these settings can confound attempts to manually trace the outlines of the calcifications with any degree of accuracy. In order to avoid this variability and the difficulty of manually tracing
May-June/1997,Volume 21, Number 3
with a trackerball, we used the density masking function to highlight the calcifications. The maximum CT value obtained previously for the circular inner ROI is the appropriate minimum value for the density masking function (and the maximum can be set at the maximum value for the scanner--say, 3000). The calcifications are then highlighted automatically, and their areas displayed (Fig. 2).
RESULTS Table 1 lists the relative size and calcium concentrations for calcifications in three of the patients studied, and the resulting values of the artherosclerosis index. All the measurements in the table are the mean of five independent estimates, obtained by different operators using the standard protocol described earlier and calculated using equation (3). They are reported as mean 4-1 SD.
Fig. 2. Calcifications are highlighted automatically using the density masking function.
Quantitative assessment of artherosclerosis * G. DOUGHERTY Several representative scans, showing a range of conditions from mild to severe calcification, are shown in Fig. 3 together with their calculated atherosclerosis indices. Values of the artherosclerosis index range from a b o u t 1 in mild cases to more than 50 in severe cases. The largest values occur with large, dense plaques. Values of the artherosclerosis index measured at the mid-L3 and mid-L4 levels are plotted against
189
patient age in Fig. 4. The overlaid lines are the linear regression line for both L3 and L4 indices together (i.e. the mean index) and the 95% confidence lines. For most (80%) patients, the value of the index at the mid-L4 level is the larger. The severity of the disease, as characterized by the value of the artherosclerosis index measured at either the L3 or L4 vertebral level, shows a significant correlation with age amongst this group. The results of linear regression (at P < 0.0001)
Fig. 3. A range of calcification conditions from mild to severe. The corresponding artherosclerotic indices were (a) 1.56+0.11 (b) 8.874-0.62 (c) 45.15-1-3.16 (units are mg cm-3).
190
Computerized Medical Imaging and Graphics
May-June/1997, Volume 21, Number 3
60 50
'tO p,
Q
0
L3 index
El
L4 index
0
fl
40
30
"6 0
20
n 10
0
40
45
~ 50
55
60
65
70
75
Age (years) Fig. 4. Artherosclerosis index, at mid-L3 (O) and mid-L4 (I-q) levels, plotted against patient age. The regression line (with 95% confidence bands) for all the data points is shown (y= -73.82 + 1.65"x: correlation coefficient (r) = 0.914 (p < 0.0001)).
for the artherosclerosis index measured at the mid-L3 level, the mid-L4 level, and the mean of these two indices with age are shown in Table 2. The F-statistic is the ratio of the mean square of the dependent variable to the mean square of the residuals. The size of the Fstatistic (and its probability value) is a guide to how important the independent variable is in explaining the behaviour of the dependent variable: a low probability value indicates that it is unlikely that such a large Fstatistic would have happened by chance. The intercept on the age axis indicates that calcifications are not likely below the age of 45 yr in patients such as these. DISCUSSION A useful indicator of the severity of artherosclerosis is the consequent arterial stenosis, defined as the percentage reduction in the stenosed lumen relative to the normal lumen. %stenosis = (AnAs As) , where An=cross-sectional area of normal arterial lumen and As=cross-sectional area of stenosed arterial lumen. In conventional aortography, the external diameter of the artery can be measured in a stenosed
region and in a normal region. It is assumed that lumen narrowing will follow the narrowing of the external wall (and therein lies an important limitation to the technique). Clinically significant disease is generally considered to correspond to a narrowing of the diameter by at least 50%, or the cross-sectional area by 75%: certainly 75% stenosis by area of the lumen has been shown to be hemodynamically significant (16-18). Computed tomography is recognized as being superior to aortography at detecting intimal calcification (19), which is a marker of artherosclerosis (20). The degree of aortic artherosclerosis has generally been appraised according to an arbitrary grading system. The condition was graded as mild, if calcification was noted at one site or if scattered focal calcifications were seen; moderate, if circumferential continuous calcifications or plaque were present at one or a few levels; and severe, if circumferential changes were present at multiple levels (11, 21). More recently a "calcium score" system has been developed and used as an indicator of coronary artery disease. The size and density of calcium deposits observed in ultrafast CT images were used to determine a semiquantitative assessment of the severity of coronary artery calcification in patients. The area of calcification (in mm 2) was multiplied by an arbitrary density score based on the peak CT
Quantitative assessment of artherosclerosis • G. DOUGHERTY
number in the identified deposits (12, 13). In the absence of a phantom to relate CT numbers to actual calcium densities, calcifications had to be defined arbitrarily as comprising groups of at least four contiguous pixels with values above 130 HU: arbitrary weightings were then assigned according to peak CT value---1 for a peak in the range 130199 HU, 2 for 200-299 HU, 3 for 300-399 HU and 4 for > 400 HU. The sum of each of these scores for all of the major epicardial vessels, notwithstanding difficulties in identifying some of them, was then taken as the CT-indicated severity for coronary artery disease. Spatial resolution (based on a 512 x 512 reconstruction matrix, with a field of view of 26 cm) was similar to conventional CT: ultrafast acquisition (~100 ms) was necessary to minimize respiratory motion. This method of assessment has a number of drawbacks, the most obvious being the reduction of the continuous variable of calcium content to an arbitrarily discretized index. A fourfold increase in mortality has been reported in a group of subjects with fluoroscopically detected coronary artery calcium (22): and a number of studies have shown a direct relationship between the extent (viz. size xdensity) of arterial calcification, albeit scored semi-quantitatively, and the severity of stenotic obstructions (11-13). Radiographically detected coronary artery calcium has been shown to predict both the pathologic findings (23, 24) and the angiographic narrowings (13, 25) associated with artherosclerotic heart disease. Calcium deposits indicate plaque formation, which in turn causes stenosis. The association persists in all age groups, but is most marked in younger patients (11). Serial studies, using the relatively low reproducibility of a calcium scoring system, have shown predictive value for future symptomatic disease and even the reduction of deposits with aggressive medical management (26). The poor reproducibility of the calcium scoring algorithm has been reported particularly for low scores which can have variabilities greater than 100% (20-27). This has important implications for screening programs, for example, for coronary disease. In order to reduce the variability in calcium scoring it was recognized that alternative algorithms were required that incorporate measurements from the entire artherosclerotic lesion rather than only the peak density, and that do not bracket wide ranges of densities into discrete groups (20). The artherosclerosis index proposed in this paper addresses both of these shortcomings, to provide a continuously variable index that is more objectively quantitative than either the calcium scoring system or visual analysis.
191
The mean calcium level in the calcifications, obtained using circular ROIs and a "shifted" histogram, can be obtained precisely and reproducibly. The method does not assume circularity of the artery or the lumen. Variability due to the selection and placement of the ROIs is small, and results in a precision of less than 1.5%. The area of the calcifications ( ~ Ai) can be determined precisely (to within +1%) using the density masking protocol, but difficulties in precisely tracking the outer margin of the aortic wall can result in errors up to -t-5% in the aortic cross-section (A). However, normalization of the calcification area by the aortic cross-section is considered important since the aortic cross-section typically increases by 20% between 42 and 72 yr for females (28). The total precision of +7% in the artherosclerosis index is superior to values obtained with calcium scoring of coronary calcifications. It allows us to distinguish many gradations of severity from mild (A/H1) to severe (AI~,50), examples of which are shown in Fig. 3. The data in Fig. 4 indicate that the proposed artherosclerosis index is capable of distinguishing continuously different levels of severity within a group of patients. The increased severity with age in our cohort is consistent with studies of coronary artery calcification (12, 28, 29) and of calcium in human aortas from autopsy (15). Generally larger values were observed for the index at the L4 level, closer to the bifurcation into the iliac arteries, compared to the L3 level, especially for the younger, less severely affected patients. This agrees with suggestions that intimal thickening and lipid accumulation, both precursors of calcification, develop mainly in regions of relatively low wall shear stress (30) such as the lateral walls of arterial junctions. Our limited data suggest that abdominal aortic calcifications are rare in Kuwaiti females less than 45 yr, but that severity increases monotonically and predictably with age after that. The coefficient of determination (r 2) assesses the proportion of the dependent variable's variability that is explained by the independent variable (with a maximum value of 1.0). The value of 0.818 (P < 0.001), obtained for the mean of the indices at L3 and L4 with age, indicates the strong age-dependence of abdominal aortic calcification. The correlation is much stronger than in other studies (e.g. r 2 = 0.303 for calcium mass vs age using plain X-rays of autopsied abdominal aortas from Canadian women(15)), probably due to the homogenous nature of our test cohort. The large values of the F-statistic (at low probability values) indicate that age is very important in explaining the values obtained for the artherosclerosis index. If
192
Computerized Medical Imaging and Graphics
future studies were to use our proposed artherosclerosis index, the rate of increase could be compared amongst different ethnic groups and populations. It has been suggested that in women, the prevalence and severity of aortic calcification closely parallels the degree of osteoporosis (30): it was postulated that as calcium is leached out of bones it is deposited in the vascular tree. The majority of our patient group (17 out of 20) were diagnosed as osteoporotic on the basis of their low vertebral bone density compared to age-matched controls. However, the correlation between their degree of osteoporosis (taken as the ratio of bone mineral densities for an age matched control and the specific patient) and their severity o f artherosclerosis (given by their artherosclerosis index) was poor (r=0.13). The corresponding value of the t-statistic of 0.578 indicates that the correlation is not significant at a probability level of 5% (viz. P < 0.05). The proposed index successfully meets the need for a reliable algorithm for assessing the severity of aortic calcifications and artherosclerosis. It may be useful in predicting the development of aortic aneurysms, which result from a weakening of the aortic wall that may be related to fragmentation of connective tissue at the boundaries of calcified plaques (32). The index is considered to be generally applicable to all arterial calcifications, and is not limited to the abdominal aorta, although precise measurement of the cross-sectional area (A) will be more difficult for tortuous vessels such as the iliac arteries. The spatial extent (~Ai) of epicardial coronary artery calcification is required for both calcium scoring and the proposed artherosclerosis index algorithm. It will generally be less than the aortic calcification presented here, and CT images of the heart (even with ultrafast acquisition) will be more prone to motion artefact. This will result in some additional loss of precision. Preliminary measurements on some ultrafast CT images of coronary artery calcification have indicated a precision of 1 0 - 1 5 % for less severe calcifications. The proposed algorithm offers the potential of monitoring the effect of various therapeutic regimes, and analyzing slight changes in the evolution of artherosclerotic disease in a patient with age if a series of CT scans are available. There is currently great concern over whether the increased calcification with age can be prevented or reversed. Whilst calcified collagenous plaques in the aorta may not be reversible once established, it is apparent that some calcification in smaller arteries can be controlled or reduced (33). With a sufficiently large group of patients, it may be possible to discriminate
May-June/1997,Volume 21, Number 3 between symptomatic and asymptomatic cases and use an individual's index to predict whether that patient might be at high risk of developing cardiovascular disease. The index could be easily modified to 3D reconstructions of plaques imaged by CT helical scanning or by MRI, by replacing the fractional areas by fractional volumes. Preliminary work has begun at three public hospitals in Kuwait using M R I images of calcified plaques to assess the prognostic significance of the volume algorithm. The algorithm may also prove useful in assessing the condition of other disease processes: it has been suggested that idiopathic retroperitoneal fibrosis (periaortitis) marc occur more frequently in patients whose aortic calcifications are more severe than expected for their age and sex (21). SUMMARY A novel quantitative index has been developed and used to quantitate the extent of aortic calcification observed incidentally in CT scans. The index was tested on a limited group of post menopausal females. It was able to distinguish continuously different levels of severity with a reproducibility of + 7 % , a performance superior to the discretized calcium score system widely used to date. The index showed a strong correlation between the severity of calcification and age (linear coefficient of correlation=0.914) with this group. It may be useful in predicting whether a patient might be at high risk of developing artherosclerotic disease (31).
Acknowledgement--Thanks are due to Dr D. Newman for
suggesting work in this area, to Ms Fatma Farhan and Ms Abeer Ibrahim for assisting with some of the measurements,and to Ms Char Hunt for making the original CT scans available. REFERENCES 1. Friedewald, S.T. Epidemiologyof cardiovascular disease. In: Cecil's textbook of medicine. J.B. Wyngaarden, L.H. Smith (eds), 18th edition. Philadelphia: Saunders; 1988:179-198. 2. Senkowsky,J.; Bell,W.H.; Kertein, M.D. Normal angiograms and carotid pathology. Am. Surg. 56:726-729; 1990. 3. Merickel,M.B.; Carman, C.S.; Brookeman, J.R.; Ayers, C.R. Image analysis and quantification of artherosclerosis using MRI. Comput. Med. Imag. Graph. 15:207-216; 1991. 4. Herfkens, R.J.; Higgins, C.B.; Hricak, H.; Lipton, M.J.; Crookes, L.E.; Sheldon, P.E.; Kaufman, L. Nuclear magnetic resonanceimaging ofartherosclerotic disease. Radiol. 48:161166; 1983. 5. Wesbey,G.E.; Higgins, C.B.; Hale, J.D.; Valk, P.E. Magnetic resonance applications in artherosclerotic vascular disease. Cardiovasc. Intervent. Rad. 8:342-350; 1986. 6. Ranke, C.; Creuzig,A.; Alexander, K. Duplex scanning of the peripheral arteries: correlation of the peak systolic velocity
Quantitative assessment of artherosclerosis • G. DOUGHERTY
7.
8.
9. I0. 11. 12.
13.
14. 15. 16.
17. 18. 19. 20.
21.
ratio with angiographic diameter reduction. Ultrasound Med. Biol. 18:433-440; 1992. Guo, Z.; Durand, L.-G.; Allard, L.; Cloutier, G.; Lee, H. Classification of lower limb arterial stenoses from Doppler blood flow signal analysis with time--frequency representation and pattern recognition techniques. Ultrasound Med. Biol. 20:335-346; 1994. Laurora, G.; Caesarone, M.R.; De Sanctis, M.T.; Belcaro, G. The ultrasonic assessment of carotid plaques in relation to cerebral events and damage over 3 years. Minerra Cardioangiol. 40:265--269; 1992. Delcker, A.; Dienere, H.C. Quantification of artherosclerotic plaques in carotid arteries by 3D ultrasound. Br. J. Radiol. 67:672-678; 1994. Sinzinger, H.; Virgolini, I. Nuclear medicine and artherosclerosis. Eur. J. Nucl. Med. 17:160-178; 1990. Siegel, C.L.; Ellis, J.H.; Korobkin, M.; Dunnick, N.R. CTdetected renal arterial calcification: correlation with renal artery stenosis in angiography. AJR 163:867-872; 1994. Agatston, A.S.; Janowitz, W.R.; Hildner, F.J.; Zusmer, N.R.; Viamonte, M.; Detrano, R. Quantification of coronary artery calcium using ultrafast computed tomography. J. Am. Coll. Card. 15:827-832; 1990. Breen, J.F.; Sheedy, P.F.; Schwartz, R.S.; Stanson, A.; Kaufmaun, R.; Moll, P.; Run, J. Coronary artery calcification detected with ultrafast CT as an indication of coronary artery disease. Radiol. 185:435-439; 1992. Weissman, G.; Weissman, S. X-ray diffraction studies of human aortic elastin residues. J. Clin. Invest. 39:1657-1666; 1960. Sherebrin, M.H.; Lim, H.L.; Roach, M. Jr. Calcium mass in human aortas from autopsy. Physiol. Meas. 15:281-289; 1994. Gould, K.L.; Lipscomb, K.; Hamilton, G.W. Physiologic basis for assessing critical coronary stenosis: instantaneous flow response and regional distribution during coronary hyperemia as measures of coronary flow reserve. Am. J. Cardiol. 33:8794; 1974. Azuma, T.; Fukushima, T. Flow patterns in stenotic blood vessel models. Biorheol. 13:337-355; 1976. Maun, F.C.; Herrick, J.F.; Essex, H.E.; Baldes, E.J. Effect on blood flow of decreasing the lumen of a blood vessel. Surg. 4:249-252; 1992. Limpert, J.D.; Vogelzang, R.L.; Yao, J.S.T. Computed tomography of aortoiliac artherosclerosis. J. Vasc. Surg. 5:814-819; 1987. DeVries, S.; Wolfkiel, C.; Shah, V.; Chomka, A.; Rich, S. Reproducibility of the measurement of coronary calcium with ultrafast computed tomography. Am. J. Cardiol. 75:973-975; 1995. Dixon, A.K.; Springall, R.G.; Kelsey, F.I.; Taylor, G.W. Computed tomography (CT) of abdominal aortic aneurysms: determination of longitudinal extent. Br. J. Surg. 68:47-50; 1981.
193
22. Margolis, J.R.; Chen, J.T.; Kong, Y.; Peter, R.H.; Behar, V.S.; Kisslo, J.A. The diagnostic and prognostic significance of coronary artery calcification--a report of 800 cases. Radiol. 137:609-615; 1980. 23. McCarthy, J.H.; Palmer, F.J. Incidence and significance of coronary artery calcification. Br. Heart J. 36; 1974:499-504. 24. Moore, E.H.; Greenberg, R.W.; Merrick, S.H.; Miller, S.W.; McLoud, T.C.; Shep, J.O. Coronary artery calcifications: significance of incidental detection in CT scans. Radiol. 172:711-716; 1989. 25. Hamby, R.I.; Tabrah, F.; Wisoff, B.G.; Hartgstein, M.L. Coronary artery calcification: clinical implications and angiographic correlates. Am. Heart J. 87:565-569; 1974. 26. Imagix, San Francisco, CA. Coronary artery disease and calcification, 1993. 27. Bielak, L.; Kaufmann, R.; Moll, P.; Mc Collough, C.; Schwartz, R.; Sheedy, P. Small lesions in the heart identified at electron beam CT: calcification or noise? Radiol. 192:631636; 1994. 28. Dixon, A.K.; Lawrence, J.P.; Mitchell, J.R.A. Age-related changes in the abdominal aorta shown by computed tomography. Clin. Rad. 35:33-37; 1984. 29. DeVries, S.; Wolfkiel, C.; Fusman, B.; Bakdash, H.; Ahmed, A.; Levy, P.; Chomka, A.; Kondos, G.; Zajac, E.; Rich, S. Influence of age and gender on the presence of coronary calcium detected by ultrafast computed tomography. J. Am. Coll. Card. 25:76-82; 1995. 30. Nerem, R.M. Vascular fluid mechanics, the arterial wall and artherosclerosis. J. Biomech. Eng. 114:274-282; 1992. 31. Boukhris, R.; Becker, K.L. Calcification of the aorta and osteoporosis. J. Am. Med. Ass. 219:1307-1312; 1972. 32. Sherebrin, M.H.; Bernans, H.A.; Roach, M.R. Extensibility changes of calcified soft tissue strips from human aorta. Can. J. Physiol. Pharmacol. 65:1878-1883; 1987. 33. Feldman, F. Soft tissue mineralization: Roentgen analysis. Current Prob. Diagnostic Radiol. 15:161-240; 1986.
About the Author-----~EOFFDOUGHERTY received a BSc from the
University of Manchester in 1971 and a PhD from Keele University in 1979. He has held faculty positions at universities in England, Switzerland, Malaysia and Australia, and was a Visiting Research Professor at Swarthmore College, Philadelphia. He is currently Professor of Medical Image Processing at the University of Kuwait. His research interests include medical applications of image processing and pattern recognition, multiresolutional texture measures, and the quantitation of global image quality. Professor Dougherty is a Senior Member of IEEE, a member of AAPM, and a Fellow of IEE and AIP (note: I E E - Institution of Electrical Engineers (UK); AIP--Australian Institute of Physics).