Videodensitometric ejection fractions from intravenous digital subtraction left ventriculograms: correlation with conventional direct contrast and radionuclide ventriculography

Relationship between the pixel value in digital subtraction angiography and iodine concentration: study in high iodine concentration with original phantom

Quantitative digital subtraction angiography (DSA) image analysis based on densitometry is widely accepted and used. For the densitometoric DSA image analysis, it is required that there is a linear relationships between the pixel values on DSA images (DSA values) and contrast medium iodine concentration or the thickness of the vessels or the chambers filled with contrast material. We studied on relationship between the DSA value and iodine concentration especially in high iodine concentration. As for the relationship between DSA values and iodine concentration on the DSA images at low concentration, DSA phantom had a good linear relationship. However, the relationship at the high iodine concentration, DSA phantom sometimes lost this linear relationship. Our results suggested that it was necessary to identify relationship between DSA values and iodine concentration in each DSA system, especially in high iodine concentration setting, for densitometoric DSA image analysis.

Digital ventriculography: valid on-line calculation of cardiac volumes by corrected computer densitometry in coronary artery disease

Cardiac catheterization facilitates the assessment of left ventricular function in coronary artery disease (CAD). Digital left ventriculography offers the potential for an on-line quantitative determination of left ventricular end-diastolic volume (LVEDV) and left ventricular end-systolic volume (LVESV). These are routinely evaluated by the area-length method (ALM), which is considered as a standard. Densitometry (DENS) is an alternative method but may lead to calculated underestimations. The purpose of this study was to investigate the accuracy of corrected DENS for LVEDV and LVESV in comparison with ALM in single-plane 30 degrees right anterior oblique (RAO) projection. The computer densitometric correction equation was adapted from the linear regression analysis (y = 0.86x + 2.73) in cardiac models and applied to the analysis of digital left ventriculograms in patients suffering from CAD. The study of cardiac models yielded highly significant correlations (r > or = 0.9; P < or = 0.001) between true volumes and both DENS and ALM. DENS and ALM correlated highly significantly (r > or = 0.9; P < or = 0.001) with a low standard error of estimate (SEE) of +/-9.5 mL. The comparison of DENS and ALM in 44 patients' digital ventriculograms exhibited highly significant (r = 0.74; P < or = 0.001) correlations for noncorrected and corrected LVEDV. Systematic underestimation by DENS of LVEDV disappeared after correction and SEE decreased from +/-23.7 to +/-19.2 mL. DENS and ALM correlated highly significantly for LVESV (r = 0.78; P < or = 0.001; SEE +/- 15.6 mL +/- 13.5 mL, respectively) after correction. Following correction, mean values for DENS increased from 116 +/- 32 to 132 +/- 37 mL (LVEDV) and 50 +/- 22 to 55 +/- 25 mL (LVESV). For ALM, mean values decreased from 159 +/- 35 to 127 +/- 28 mL (LVEDV) and 55 +/- 25 to 46 +/- 21 mL (LVESV). This study shows that LVEDV and LVESV can be reliably analyzed on-line by corrected computer densitometry. Method-related errors of both DENS and ALM are present and account for minor volume deviations.

On-line evaluation of systolic performance by densitometry in digital left ventriculography

The angiocardiographic evaluation of left ventricular end-diastolic (LVEDV) and end-systolic (LVESV) volumes and ejection fraction (EF) is routinely performed by the area-length method (ALM) but may lead to erroneous results. Digital imaging in real time allows densitometric procedures of determining left ventricular (LV) performance to be applied alternatively. In this study, we present densitometric algorithms for the analysis of LVEDV, LVESV, and EF from digital image data, establish accuracy and reproducibility, and determine value and limitations in comparison with ALM in single-plane 30 degrees right anterior oblique (RAO) projection. A linear relationship between iodine depth and measured densities is mainly burdened with scatter radiation and beam hardening which reduce primary radiation and suppress iodine depth. However, facilities such as deconvolution and correction algorithms are capable of reducing these sources of error. In the present study, computer-analyzed contrast images of iodine-filled wedges and spheres showed a near-linear relationship between iodine depth between 50-100 mg/cm2 and measured densities. Contrast images of heart casts and LV angio-grams of 54 patients were obtained with a digital image acquisition and processing system, and evaluated by two independent observers. The phantom study resulted in significantly (p < or = 0.01) better densitometric standard errors of estimate for volumes [3.3 ml densitometry (DENS) vs. 8.9 ml (ALM)] and simulated EF [4.3% (DENS) vs. 7.8% (ALM)] than ALM. The standard error of estimate for the comparison between both methods was 8.4 ml for volumes and 7.5% for EF. Densitometric volumes tended to underestimate volumes calculated by ALM. The angiographic study of patients demonstrated significant correlations between both methods (LVEDV r = 0.78, LVESV r = 0.83, total volumes: r = 0.89; EF r = 0.88). The standard errors of estimate can be ascribed to systematic, method-related errors of both DENS and ALM (LVEDV +/- 28.9 ml, LVESV +/- 23.4 ml, total volumes (EDV and ESV) +/- 27.1 ml; EF +/- 8.1%). The intra- and interobserver variability, respectively, exhibited significantly smaller (p < or = 0.01 and p < or = 0.05, respectively) standard errors of estimate for densitometric EF [4.6% (DENS) vs. 8.5% (ALM) and 7.1% (DENS) vs. 10.3% (ALM), respectively]. Inclined but not significant differences were found for LVEDV and LVESV. In conclusion, the data presented indicate that the calculation of LV volumes and EF in digital left ventriculography may be performed accurately by densitometric calculation in single-plane 30 degrees RAO projection. Minor underestimations in densitometric volume determination may be anticipated in the evaluation of LV geometry.

Improved densitometric measurement of left ventricular ejection fraction using dual-energy digital subtraction angiography

The effects of misregistration artifacts and background corrections on the densitometric measurement of left ventricular ejection fraction (EF) from digital subtraction angiography (DSA) images were studied in 20 patients. Densitometric ejection fraction measurements were performed on both conventional time subtraction images and on dual-energy subtraction images. Dual-energy subtraction is not sensitive to the motion induced artifacts which often mar time subtraction images. While the time subtraction images had varying degrees of misregistration artifacts, none of the dual-energy studies contained significant misregistration artifacts. Densitometrically determined ejection fractions measured with and without correction for background signals were compared. Poor agreement between time subtraction ejection fractions determined with and without background correction (EFNO-BKG = 0.88 EFBKG - 6.0%, SEE = 8.1%, r = 0.83) demonstrated the sensitivity of time subtraction EFs to the performance of a background correction procedure. Conversely, densitometric measurement of ejection fraction using dual-energy subtraction was significantly less sensitive to the performance of a background correction (EFNO-BKG = 0.99 EFBKG - 5.3%, SEE = 4.3%, r = 0.96). Since background correction requires accurate definition of ventricular borders, but motion artifacts often preclude accurate border definition, it is concluded that dual-energy subtraction is a significantly more robust method for measuring left ventricular ejection fraction using densitometry. It is further concluded that identification of the systolic endocardial border is not required when performing densitometric EF measurements on dual-energy images. Drawing of the end-diastolic border alone is sufficient to produce an accurate ejection fraction measurement.

Densitometric assessment of regional left ventricular systolic function during graded ischemia in the dog by use of dual-energy digital subtraction ventriculography

Densitometric analysis of images obtained by digital subtraction angiography (DSA) allows for more reproducible and less operator-dependent quantitation of ventricular function. Conventional DSA uses temporal subtraction but is limited by misregistration artifacts. Dual-energy digital subtraction angiography (DE-DSA) is immune to such misregistration artifacts. The ability of DE-DSA to quantitate changes in regional ventricular volume resulting from ischemia was tested. Densitometric analysis of both phase-matched and ejection fraction DE-DSA images was used to quantitate regional left ventricular systolic function during four levels of ischemia ranging from mild to severe in open-chest dogs (n = 10). DE-DSA left ventriculograms were obtained by means of central venous injections of iodinated contrast medium. Ischemia was graded according to percentage of systolic wall thickening as measured by sonomicrometry. Phase-matched end-systolic images were obtained at each of four levels of ischemia by subtracting an end-systolic control image from each end-systolic ischemic image. Ejection fraction images were obtained at the control level and at each level of ischemia by subtracting an end-systolic image from an end-diastolic image of the same cardiac cycle. The resulting wall motion difference signals represent the changes in regional ventricular volumes and were quantitated by densitometry. Densitometry was able to detect the effect of all levels of ischemia on regional function, even the mildest. Densitometric analysis of both phase-matched and ejection fraction DE-DSA images provides a sensitive technique for detecting and quantitating the changes in regional left ventricular systolic volume that occur with ischemia.

Time-course of left atrial performance during coronary artery occlusion followed by reperfusion in anesthetized dogs by densitometric analysis of digital atrioventriculographic images

The left atrial (LA) function during coronary artery occlusion followed by reperfusion using densitometric analysis of digital atrioventriculographic images was evaluated. Eight anesthetized dogs underwent atrioventriculography at baseline, 10 and 60 min after left circumflex coronary artery (LCX) occlusion and 5, 30, 60, and 120 min of reperfusion. Time-density curves were obtained for LA and left ventricle (LV). The ratios of passive atrial video-densitometric change (VC) to total VC (Passive Ratio), and active VC to total VC (Active Ratio) were calculated. Left ventricular ejection fraction (LVEF), peak ejection rate (PER), and peak filling rate (PFR) were derived. Active Ratio, an index of atrial contraction, increased to 144%, and Passive Ratio decreased to 75% of baseline at 60 min of LCX occlusion. Two hours after reperfusion, both Active and Passive Ratios returned to control level. While LVEF reduced to 70%, PER to 67%, LV peak positive dP/dt to 88% of baseline at 60 min after occlusion, and remained depressed at 2 h after reperfusion. However, PFR, LV peak negative dP/dt and LV isovolumic pressure decay rate showed recovery at 2 h after reperfusion. There were significant correlations between PFR and Passive Ratio (r = 0.41), and between Active and Passive Ratios (r = 0.55). Thus, time-course of recovery of LV post-ischemic systolic and diastolic function was different. Return of LA function to control level during 2 h after reperfusion may be depend on recovery of LV diastolic function.

A densitometric method for quantitative analysis of the left ventricle performance using i.v. digital subtraction angiography

In this paper we present a method we used to provide quantitative description of the systolic and diastolic temporal function of the left ventricle (LV). Additional parameters, such as peak filling and ejection rates, times to end systole and end diastole, and temporal changes in slow and fast filling are obtained. The volumes associated with these parameters are also calculated. Correlation between LV volume changes during the cardiac style and corresponding "density" variations was confirmed. Time-density curves were obtained from selected cardiac cycles in each study. We used the polynomial fitting technique to fit the time density curves and developed a computer algorithm for deriving the relevant parameters. Data from a total of 18 patients with ischemic heart or valvular diseases, who underwent I.V. ventriculography was analysed using our method. Some of these patients were forwarded for repeated digital subtraction angiography (DSA) examination before and after intervention therapy to evaluate the effectiveness of treatment. In comparison to the geometric method for the analysis of LV performance, our method is generally faster and simpler to employ. The method was effective in detecting variations in the peak ejection and filling rates in our group of patients before and after interventional therapy.

Determination of cardiac ejection and valvular regurgitant fraction by on-line digital densitometry--methodology, validation, and application

Videodensitometry allows to obtain both left and right ventricular ejection fraction (EF) and aortic or pulmonary regurgitant fraction (RGF) from the wash-out curve of contrast medium. We developed this technique to digital densitometry and integrated it in the standard digital image acquisition system 'Digitron' using Siemens user's library. Sources of error like scatter radiation, veiling glare, accumulation of iodine in tissue, and inhomogeneous contrast mixing were considered by using ECG gated image subtraction, background reference regions, data fit to ideal wash-out curves and calculation of EF and RGF exclusively from density differences. The method was validated by phantom studies in which simulated angiocardiograms were generated with given values of EF (50 to 70%) and RGF (0 to 45%). The results tended to overestimate RGF by up to 10 percent points, when image contrast was high and the ventricle was masked poorly by the lead shutters. In the clinical setting, the reliability of the results can be judged from the fit of the wash-out curve presented automatically on the screen on a semi-logarithmic scale. The technique is available to the physician in the catheterization laboratory on-line during or immediately after the examination, which facilitates routine use.

Evaluation of the densitometric method of ejection fraction by direct digital subtraction left ventriculography

The aim of the present study was to test the hypothesis that the densitometric ejection fraction (EF) could be calculated by direct contrast digital subtraction left ventriculography in a single heartbeat. Thirty patients underwent direct contrast digital subtraction left ventriculography and biplane conventional left ventriculography. The values of EF were obtained from digital subtraction ventriculograms of the first cardiac cycle after completion of the injection of the contrast medium by the videodensitometric technique and were compared with those obtained from conventional ventriculograms by the area-length method in the same patients. There was a close correlation between the densitometric EF from digital subtraction ventriculograms and the volumetric EF from conventional ventriculograms (r = 0.78). Densitometric EF measured in the 30 degree right anterior oblique projection highly correlated with densitometric EF measured in the 60 degree left anterior projection (r = 0.98). These data suggest that this densitometric technique is available in any projection and is a simple procedure for the accurate measurement of left ventricular EF.

Left ventricular dual-energy digital subtraction angiography: a motion immune digital subtraction technique

Digital subtraction angiography (DSA) allows quantitative analysis of ventricular function via densitometric and parametric imaging techniques. However, DSA is limited by the artifacts in temporal subtraction images that result from patient and cardiac motion. Dual-energy subtraction imaging is insensitive to motion. This study evaluated the initial application of dual-energy subtraction in cardiac patients. The image quality of dual-energy subtraction left ventriculograms obtained from a pulmonary artery injection of contrast was assessed in 13 patients, ranging in weight from 54 to 100 kg. The dual-energy images were compared with left ventricular images obtained using standard left ventricular injection cine angiography. End-systolic and end-diastolic ventricular volumes calculated from the cine (C) and dual-energy (DE) images using the Area-Length method were compared. The resulting regression line was DE = 0.98 C+ 7.0 ml, and the r value was 0.987. Dual-energy subtraction provided good left ventricular visualization, free from misregistration artifacts, even during patient motion.

Limitations of the lead oxide vidicon for dual-energy digital subtraction angiography

Dual-energy digital subtraction angiography (DSA) can be performed, with some modifications, using existing systems designed for conventional DSA work. However, the video camera generally found in such installations, the lead oxide vidicon, places significant limitations on the image quality of dual-energy DSA. If the video camera is operated in a 60 Hz, 266 line progressive scanning mode to maintain a nominal 30 subtraction images/s frame rate, the maximum available X-ray pulse width is limited to 5 ms. The resulting required increase in low-energy kVp reduces the dual-energy signal-to-noise ratio by approximately 40% for cardiac dual-energy DSA imaging. In addition, the performance of the video camera with respect to read-out lag is much more important for dual-energy DSA than conventional single-energy imaging. Measurements show that lag reduces the iodine contrast in dual-energy DSA by up to 20% when a 30 Hz vertical scan mode is used, and by up to 27% when a 60 Hz vertical scan mode is employed. Replacement of the lead oxide vidicon with a CCD camera removes these limitations.

A clinical application of iodine quantitation using digital fluorography

A preliminary clinical study of 20 patients is described in which a method of iodine quantitation in vitro using digital fluoroscopy was applied to the intravenous urogram. It was possible to quantitate iodine levels, in terms of mass thickness, in both the renal parenchyma and collecting system in 38 of 40 kidneys studied (20 patients). In the two kidneys on which measurements could not be made, the cause of failure was subtraction artefact due to movement of bowel gas, compounded in one case by poor renal opacification. The variations of iodine mass thickness against time were plotted and their forms were similar to dynamic computed tomography and magnetic resonance imaging studies of transplanted kidneys. Potential clinical applications of this method, in particular the functional assessment of the dilated upper urinary tract, are suggested based on its ability to offer quantitative information in addition to anatomical images.

Evaluation of magnetic resonance imaging for determination of left ventricular ejection fraction and comparison with angiography

Left ventricular ejection fraction was measured by magnetic resonance imaging (MRI) and compared with standard monoplane left ventriculography in 46 patients with various cardiac diseases. Two different MRI strategies were used. In 28 patients (group 1), ejection fraction was determined using a single slice comparable with the right anterior oblique projection of the ventriculogram. Comparison of left ventricular ejection fraction yielded a poor correlation between single slice MRI (y) and ventriculography (x) (y = 28.7 + 0.47 x, r = 0.65). In 18 patients (group 2), a multiple contiguous slice MRI technique was used to allow ejection fraction and stroke volume determination by summing up the volumes of ventricular cavity intersections. Regression analysis showed a high correlation between multiple slice MRI (y) and ventriculography (x) (y = 7.2 + 0.88 x, r = 0.98). Also, correlation between MRI right (y) and left (x) ventricular stroke volumes was satisfactory, (y = -12.8 + 1.09 x, r = 0.83). It is concluded that the multiple slice imaging technique in MRI provides an accurate noninvasive means for quantification of left ventricular ejection fraction that can be extended to the determination of left ventricular volume.

Quantitative assessment of regional left ventricular function by densitometric analysis of digital-subtraction ventriculograms: correlation with myocardial systolic shortening in dogs

Conventional wall motion analysis of contrast ventriculograms assesses only that part of the wall that is tangential to the x-ray beam. To assess regional left ventricular function in three dimensions, a new computerized method based on densitometric analysis of digital subtraction left ventriculograms was developed and validated in nine open-chest dogs instrumented with a circumflex coronary artery occluder and sonomicrometers in the anterior and posterior walls. Each dog underwent digital subtraction ventriculography at baseline and at five levels (I to V) of dysfunction of the inferior wall induced by progressive stenoses of the circumflex coronary artery. The ventriculogram was divided into six segments around the end-diastolic center of gravity. Time-volume curves were obtained by densitometry in the normal anterior and ischemic inferior segments containing the sonomicrometers. From these curves, regional ejection fraction (R-EF), regional peak ejection rate (R-PER), and regional phase (R-PH) and amplitude (R-AMP) of the first Fourier harmonic were derived. From baseline to level V of dysfunction, myocardial systolic shortening determined by sonomicrometry decreased by 124 +/- 34% of control (mean +/- SD; p less than .001) in the ischemic wall, while it increased by 12 +/- 19% (NS) in the normal wall. At the same time, R-EF, R-PER, and R-AMP decreased in the ischemic segment by 65 +/- 12%, 46 +/- 30%, and 45 +/- 15% of control, respectively (all p less than .01), while they remained unchanged or increased in the normal segment. R-PH was delayed by 14 +/- 5% (p less than .01) in the ischemic segment, but remained unchanged in the normal segment, reflecting the asynchrony of regional left ventricular contraction during ischemia. Densitometric indexes of regional function correlated well with sonomicrometric systolic shortening both in normal and ischemic segments, with r values of .84 for R-EF, .80 for R-AMP, .64 for R-PER, and .55 for R-PH (all p less than .0001). Thus, densitometric analysis of digital subtraction left ventriculograms allows three-dimensional assessment of the extent, velocity, and synchrony of regional left ventricular contraction. Densitometric indexes of regional contraction correlate well with direct measurements of myocardial systolic shortening and are useful in quantitating regional left ventricular dysfunction.

Densitometric regional ejection fraction: a new three-dimensional index of regional left ventricular function--comparison with geometric methods

Densitometric regional ejection fraction obtained by computer analysis of digital subtraction ventriculography was evaluated as a new, quantitative, three-dimensional index of regional left ventricular performance. Eighteen patients with coronary artery disease and seven control subjects had right anterior oblique ventriculography at rest and immediately after rapid atrial pacing using central venous injection of contrast material. Regional left ventricular ejection fraction was determined by densitometry in six segments drawn around the end-diastolic center of gravity, and compared with two conventional indexes of segmental wall motion: area and radial regional ejection fraction. Densitometric, area or radial regional ejection fraction was classified as abnormal if it fell at least 2 standard deviations below the corresponding mean value in the normal group. The densitometric method did not require outlining of the end-systolic left ventricular silhouette and was the easiest and fastest to perform of all three techniques. In addition, intra- and interobserver reproducibilities were higher with the densitometric method (r = 0.97 and 0.95) than with either the area (r = 0.84 and 0.82) or the radial method (r = 0.82 and 0.76). Regional left ventricular dysfunction as assessed by the densitometric, area and radial techniques allowed the detection of coronary artery disease in 50, 50 and 44% of the patients at rest and in 83, 67 and 61% of the patients in the post-pacing period, respectively. Post-pacing regional left ventricular dysfunction accurately predicted the presence or absence of greater than 70% diameter stenosis in the supplying coronary artery in 75, 67 and 56% of the cases, respectively. Thus, densitometric analysis of digital subtraction ventriculography allows a fast and reproducible three-dimensional determination of regional left ventricular ejection fraction. Using this technique, pacing-induced regional dysfunction can be detected in most patients with coronary artery disease and corresponds well with the location of significant coronary artery lesions.

Validation of a digital videodensitometric program analysis for measurement of left ventricular ejection fraction

Left ventricular (LV) function was studied in 30 patients using digital subtraction angiography by the intravenous approach. Each ventriculogram was processed with a specific videodensitometric analysis to determine LV ejection fraction. The program was verified in an experimental set-up consisting of nine latex balloons filled with contrast medium. Its validation has been established by comparing videodensitometric results with classical results supplied by geometric methods. A good correlation was obtained (r = 0.9449) and, furthermore, with experimental models, videodensitometric analysis seemed to be more accurate than geometric analysis. Digital videodensitometry appears to be a valuable and accurate method for quantifying LV function, and a promising technique for determination of the real volumes.

Videodensitometric ejection fraction from intravenous digital subtraction right ventriculograms: correlation with first pass radionuclide ejection fraction

Thirty-one consecutive patients undergoing intravenous blurred mask digital subtraction right ventriculography were submitted to first pass radionuclide angiography. Second order mask resubtraction of end-diastolic and end-systolic right ventricular digital image frames was executed using preinjection end-diastolic and end-systolic frames to rid the digital subtraction images of mis-registration artifact. End-diastolic and end-systolic perimeters were drawn manually by two independent observers with a light pen. Ejection fractions calculated from the integrated videodensitometric counts within these perimeters correlated well with those derived from the first pass radionuclide right ventriculogram (r = 0.84) and the interobserver correlation was acceptable (r = 0.91). Interobserver differences occurred more frequently in patients with atrial fibrillation and in those whose tricuspid valve planes were difficult to discern on the digital subtraction right ventriculograms. These results suggest that videodensitometric analysis of digital subtraction right ventriculograms is an accurate method of determining right ventricular ejection fraction and may find wide clinical applicability.