Hcj Aim*" Vic AJH

V"nL E^flt

I litiv

FIGURE 1 Left ventricular volume calculated from cine-MRI using the area-length geometric method applied to two orthogonal planes.

distortions of the X-ray beam geometry, and veiling glare [2123]. Moreover, these factors are not uniform over the field of view and therefore cannot be corrected by measurements performed in a single location. Variation in the overlaying noncardiac structure attenuation further complicates these problems. Inhomogeneous distribution of contrast within the cardiac chambers can also be a significant problem. Despite these difficulties, accurate estimation of ejection fractions has been reported [24-26].

The densitometric analysis is comparable to count-based radionuclides in their evaluation of ejection fraction. For the calculation of the ventricular ejection fraction, regions of interest containing the ventricular chamber at end-diastole and end-systole are extracted from the images. A third region of interest representing the background is usually obtained at end-systole forming a U-shaped area between the end-diastolic and end-systolic region of interest (Fig. 2).

The ejection fraction is then calculated by the formula

where EDd equals the summed intensity in the end-diastolic area, ESd the intensity in the end-systolic, and BKd the background intensity. To obtain total background intensity, the average background pixel value is extrapolated for the total number of pixels in end-diastole:

Background intensity at ES Number of pixels in BK

necessary to trace the borders of the ventricular end-systole and diastole, no geometric assumptions regarding left ventricular shape are made.

In a clinical evaluation study reported by Nissen etal. [27], the densitometric ejection fraction correlated closely to both the single-plane RAO and biplane area-length methods for a group of 72 patients who presented with a wide variety of cardiac disorders and ejection fraction ranging from 24% to 82%. The correlation was equally close for the subgroups of patients in whom digital ventriculography was performed following either intravenous or direct ventricular injection. Accordingly, these data indicate that calculation of ejection fraction by densito-metry can be applied to digital subtraction images produced by either site of injection. In the subgroup of patients without prior

FIGURE 2 Schematic representation of end-diastolic and end-systolic regions of interest. The background region is the area between end-diastole and end-systole.

This algorithm is very similar to that utilized in the calculation of ejection fraction by radionuclide methods. Although it is

FIGURE 2 Schematic representation of end-diastolic and end-systolic regions of interest. The background region is the area between end-diastole and end-systole.

myocardial infarction, densitometric analysis of ejection fraction correlated closely with both single-plane RAO and biplane area-length methods. Presumably, left ventricular geometry in these patients conforms to the geometric ellipsoid assumption of the area-length method. It is therefore not surprising that densitometry correlates nearly as well with single plane RAO cineangiography as biplane cineangiography in this noninfarc-tion group. Significantly, however, patients who had previously suffered myocardial infarction were studied and ejection fraction correlated more closely to the biplane than the singleplane area-length method. These data strongly suggest that densitometric ejection fraction obtained from a single RAO view may reflect global left ventricular function more accurately than single-plane area-length methods in patients with dyssynergy. Many catheterization laboratories lack facilities for biplane angiography, and densitometry analysis appears to be preferable to area-length methods for the assessment of ejection fraction in this setting.

Several practical factors need to be considered in the technique employed for densitometry analysis. Densitometric assessment of ejection fraction relies upon a linear relationship between the concentration of contrast and optical brightness. However, a variety of nonlinearities are involved in the process. These include the possibility of a nonlinear response from the image intensifier or fluoroscopic camera that may be present in some X-ray systems. Many X-ray systems exhibit vignetting, a phenomenon that causes a fall-off in image brightness from the center to the edge of the field. Geometric distortions due to the pincushion effect also can be a potential nonlinearity. Furthermore, radiographic scatter and optical veiling glare can result in a nonlinear relationship between contrast concentration and pixel gray value. Registration artifacts due to patient motion are an additional concern. It was reported that as many as 20% of patients are unable to hold inspiration and to remain motion-free for sufficient time to perform a study [5]. When a registration artifact is present due to misalignment of images before and after contrast injection, calculation of ejection fraction by densitometry should be avoided. If densitometry is to be performed following direct LV or RV injection of contrast, several additional cautions are appropriate. All cardiac cycles during the injection of contrast cannot be utilized since additional contrast is continuously entering the cavity. Because of concerns regarding mixing, most investigators have limited studies to intravenous contrast administration.

both spatial features and motion information in identifying various cardiac structures. Most quantitative analyses of gated blood pool images require the calculation of one or more regional time activity curves. Analysis of left ventricular time activity curves can yield several measures of left ventricular systolic performance, including the global ejection fraction peak ejection rate and various temporal measures [5,28-30].

Left ventricular ejection fraction is calculated from the counts in the left ventricular region of interest at end-diastole and end-systole using the formula

where ED- and ES- are the background corrected end-diastolic and end-systolic counts. Background correction is necessary since about 30% of the activity in the left ventricle is contributed by noncardiac structures. The procedure for determining the ejection fraction essentially consists of identifying regions of interest that contain the left ventricle at end-diastole and end-systole and one or more appropriate background regions.

Following identification of end-diastolic and end-systolic regions of interest, the background region of interest can be automatically identified on the end-systolic frame. It is usually located to the lower right and follows the contour of the left ventricle. In general the background region of interest will fall in a position in the image that is partially within the end-diastolic region of interest. The rationale for this approach is that it attempts to measure background activity at the position that is superimposed on the left ventricle during diastole and thus, is more representative of the true ventricular background. The ejection fraction can be determined from the region of interest data using the complete formula

ED - Bed where ED and ES are the end-diastolic and end-systolic counts and Bed and Bes are the corresponding background correction factors. The accuracy and reproducibility of left ventricular ejection fraction determined by gated blood pool imaging depend upon careful selection of background and ventricular regions of interest. This is particularly critical in patients with very high (> 80%) or very low (< 20%) ejection fractions [31]. For this reason, most institutions usually report such patients as having an ejection fraction above 80% or less than 20%.

Radionuclide Blood Pool Images

Both qualitative and quantitative approaches have been employed for the evaluation of gated radionuclide angiograms, also referred to as "blood pool" studies. Visual interpretation of blood pool images can best be accomplished by displaying the image sequence as an endless loop movie using a computer-based display. Such a movie display allows the observer to use

0 0

Post a comment