To provide an approach suitable for on-line analysis of ventricular function, a conventional two-dimensional ultrasound imaging system was modified to detect and track blood-tissue interfaces in real time based on their quantitative acoustic properties. This modification permitted on-line display of the left ventricular cavity area, fractional area change, volumes and ejection fraction on a beat by beat basis. Images were obtained from 54 patients and 12 normal subjects with broad ranges of ventricular dimensions and systolic function. On-line measurements of cavity areas were compared with off-line measurements of cavity areas (analysis of videotaped conventional images). Left ventricular cavity areas measured on-line from short-axis views correlated closely with off-line views as did areas from apical views. On-line fractional area change correlated well with ejection fraction calculated off-line. More than 70% of patients could be studied adequately with the approach developed. Thus, automatic boundary detection based on quantitative assessment of tissue acoustic properties permits on-line quantitation of ventricular cavity areas and indexes of function.

Real-time on-line automated edge detection provides accurate estimation of manually drawn cavity areas. Although the method is gain dependent, measurements are reproducible. The system should have clinical application in settings in which measurements of left ventricular function are important.

We have shown previously that the physical properties of myocardium in dogs can be characterized with quantitative ultrasonic integrated backscatter and that interrogation of the tissue with ultrasound can delineate cardiac cycle-dependent changes in ultrasonic backscatter in normal tissue that disappear with ischemia and reappear with reperfusion if functional integrity is restorable. To determine whether this approach can be applied to man, we implemented an automatic gain compensation and continuous data acquisition system to characterize myocardium with quantitative ultrasonic backscatter and to detect cardiac cycle-dependent changes in real time. We developed a two-dimensional echocardiographic system with quantitative integrated backscatter imaging capabilities for use in human subjects that can automatically differentiate ultrasonic signals from blood as opposed to those obtained from tissue and adjust the slope of the gain compensation appropriately. Real-time images were formed from a continuous signal proportional to the logarithm of the integrated backscatter along each A-line. In our initial investigation, 15 normal volunteers (ages 17 to 40 years, heart rates 44 to 88 beats/min) and five patients with dilated cardiomyopathy (ages 22 to 52, heart rates 82 to 120 beats/min) were studied with conventional parasternal long-axis echocardiographic views. Diastolic-to-systolic variation of integrated backscatter in the interventricular septum and left ventricular posterior wall was seen in each of the normal subjects averaging 4.6 + 1.4 dB (SD) and 5.3 + 1.5 dB (n = 127 sites), respectively. In patients with dilated cardiomyopathy, the magnitude of this variation was either reduced or absent, averaging 0.9 + 0.8 dB in the interventricular septum and 1.8 + 1.2 dB in the left ventricular posterior wall (n = 31 sites; p < .01 for both). in eight of the 31 sites in myopathic hearts, no variation was detectable. The results obtained demonstrate that quantitative ultrasonic tissue characterization is feasible in man. Real-time integrated backscatter imaging delineates cardiac cycle-dependent changes in normal human myocardium and quantitatively differentiates between normal and myopathic myocardium. This system offers promise for the quantitative, diagnostic detection of diverse disease processes, including myocardial ischemia and responses of the tissue to reperfusion.

Integrated ultrasonic backscatter has been related to collagen deposition in fibrotic myocardium. The purpose of our study was to measure the integrated ultrasonic backscatter in the right and left ventricles of 10 normal freshly excised canine hearts and five normal formalin-fixed human hearts. A 2.25 MHz, 50% fractional bandwidth transducer was positioned at the transducer focal distance from the epicardium. The radio frequency backscatter signal, excluding specular reflections, was digitized, squared, and integrated to yield the integrated ultrasonic backscatter (in decibels down from a 100% reflector). The segment of myocardium corresponding to the integrated ultrasonic backscatter sample volume was excised and assayed for hydroxyproline, a marker for collagen. A second purpose of our study was to evaluate the influence of fixation with formalin on the backscatter. Regional integrated ultrasonic backscatter was therefore measured in 10 freshly excised canine left ventricles, which were fixed in 10% formalin for 2 weeks. Integrated ultrasonic backscatter measurements were then repeated. In freshly excised canine hearts, the integrated ultrasonic backscatter from right ventricle was higher than that from left ventricle (-60.4 ± 1.6 [SEMI vs -66.9 + 1.0 dB; p < .001). The collagen content of right ventricle was also higher than that of left ventricle (4.40 ± 0.26 [SEMI vs 3.58 ± 0.13 ,g/mg dry weight; p < .005). Similar results were obtained in human hearts. There were no correlations between integrated ultrasonic backscatter and collagen content (r = .28 and .32 for dogs and humans, respectively). The integrated ultrasonic backscatter in freshly excised canine left ventricles was -64.8 ± 0.7 dB, which was not significantly different than the value of -65.6 + 0.7 dB measured after fixation with formalin. We conclude that both integrated ultrasonic backscatter levels and collagen content are higher in the right ventricle than in the left ventricle of normal hearts. Formalin is a suitable fixative for ultrasound studies of the myocardium in vitro. Circulation 69, No. 4, 775-782, 1984. BOTH the importance and limitations of M mode and two-dimensional echocardiography are well known to clinicians.' Current clinical applications of echocardiography depend on specular reflection occurring at tissue interfaces to depict such structures as the epicardium, endocardium, and valve leaflets; standard echocardiographic examinations are not designed to evaluate the myocardium itself. This limitation has prompted interest in other acoustic variables of potential clinical relevance for characterization of structural changes in the myocardium. These HOYT et al. ed that collagen may be an important determinant of backscatter. These investigators found significantly increased backscatter in the hearts of rabbits and dogs 5 to 16 weeks after infarction and in the hearts of rabbit after cardiotoxic doses of doxorubicin.5 9 In both lesions these changes were associated with an increase in the collagen content based...

SUMMARY. We hypothesized that acute myocardial infarction could be detected in standard twodimensional echocardiograms of closed-chest dogs by evaluating regional echo amplitude distributions using computerized image analysis. We tested this hypothesis by performing standard, 2.4 MHz two-dimensional echoes before and 2 days after circumflex coronary occlusion in seven closedchest dogs. Control and infarcted regions of interest were studied in digitized stop-frame images. Average gray level was calculated for each region of interest, and the shape of the gray-level distribution was analyzed by calculation of skewness and kurtosis and by qualitative features of shape. Average gray level increased significantly from the pre-to postocclusion images in the infarcted regions (16.7 ± 4.2 vs. 32.4 ± 4.4 units, P < 0.01), but not in the control regions (17.4 ± 4 vs. 22.3 ± 5.5., P = NS). Average gray level could not distinguish between infarcted and normal regions within the postocclusion images (36 ± 5.2 vs. 33.6 ± 5.8, P = NS). Three independent observers qualitatively evaluated histogram shape and correctly identified 7/7 MI regions (100% sensitivity) and 14/20 normal regions (70% specificity). Quantitatively, infarct regions exhibited a significant decrease in kurtosis (2.8 ± 0.9 to 0.44 ± 0.5, P < 0.01); the normal regions showed no significant change in kurtosis from pre-to postocclusion images (7.1 ± 4.0 vs. 5.2 ± 2.9, P = NS). Within postocclusion images, infarcted regions displayed a significantly lower kurtosis than did normal regions (0.27 ± .47 vs. 2.5 ± 1.0, P < .01). We conclude that acute myocardial infarction may be detected in closed-chest dogs by analyzing regional echo amplitude data from standard twodimensional echocardiograms (Circ Res 52: 36-44, 1983)