Gerold Porenta

Predictors of Outcome in Severe, Asymptomatic Aortic Stenosis

BACKGROUND Whether to perform valve replacement in patients with asymptomatic but severe aortic stenosis is controversial. Therefore, we studied the natural history of this condition to identify predictors of outcome. METHODS During 1994, we identified 128 consecutive patients with asymptomatic, severe aortic stenosis (59 women and 69 men; mean [+/-SD] age, 60+/-18 years; aortic-jet velocity, 5.0+/-0.6 m per second). The patients were prospectively followed until 1998. RESULTS Follow-up information was available for 126 patients (98 percent) for a mean of 22+/-18 months. Event-free survival, with the end point defined as death (8 patients) or valve replacement necessitated by the development of symptoms (59 patients), was 67+/-5 percent at one year, 56+/-5 percent at two years, and 33+/-5 percent at four years. Five of the six deaths from cardiac disease were preceded by symptoms. According to multivariate analysis, only the extent of aortic-valve calcification was an independent predictor of outcome, whereas age, sex, and the presence or absence of coronary artery disease, hypertension, diabetes, and hypercholesterolemia were not. Event-free survival for patients with no or mild valvular calcification was 92+/-5 percent at one year, 84+/-8 percent at two years, and 75+/-9 percent at four years, as compared with 60+/-6 percent, 47+/-6 percent, and 20+/-5 percent, respectively, for those with moderate or severe calcification. The rate of progression of stenosis, as reflected by the aortic-jet velocity, was significantly higher in patients who had cardiac events (0.45+/-0.38 m per second per year) than those who did not have cardiac events (0.14+/-0.18 m per second per year, P<0.001), and the rate of progression of stenosis provided useful prognostic information. Of the patients with moderately or severely calcified aortic valves whose aortic-jet velocity increased by 0.3 m per second or more within one year, 79 percent underwent surgery or died within two years of the observed increase. CONCLUSIONS In asymptomatic patients with aortic stenosis, it appears to be relatively safe to delay surgery until symptoms develop. However, outcomes vary widely. The presence of moderate or severe valvular calcification, together with a rapid increase in aortic-jet velocity, identifies patients with a very poor prognosis. These patients should be considered for early valve replacement rather than have surgery delayed until symptoms develop.

Improved assessment of mitral valve stenosis by volumetric real-time three-dimensional echocardiography

OBJECTIVES This study was performed to determine the feasibility, accuracy and reproducibility of real-time volumetric three-dimensional echocardiography (3-D echo) for the estimation of mitral valve area in patients with mitral valve stenosis. BACKGROUND Planimetry of the mitral valve area (MVA) by two-dimensional echocardiography (2-D echo) requires a favorable parasternal acoustic window and depends on operator skill. Transthoracic volumetric 3-D echo allows reconstruction of multiple 2-D planes in any desired orientation and is not limited to parasternal acquisition, and could thus enhance the accuracy and feasibility of calculating MVA. METHODS In 48 patients with mitral stenosis (40 women; mean age 61 +/- 13 years) MVA was determined by planimetry using volumetric 3-D echo and compared with measurements obtained by 2-D echo and Doppler pressure half-time (PHT). All measurements were performed by two independent observers. Volumetric data were acquired from an apical view. RESULTS Although 2-D echo allowed planimetry of the mitral valve in 43 of 48 patients (89%), calculation of the MVA was possible in all patients when 3-D echo was used. Mitral valve area by 3-D echo correlated well with MVA by 2-D echo (r = 0.93, mean difference, 0.09 +/- 0.14 cm2) and by PHT (r = 0.87, mean difference, 0.16 +/- 0.19 cm2). Interobserver variability was significantly less for 3-D echo than for 2-D echo (SD 0.08cm2 versus SD 0.23cm2, p < 0.001). Furthermore, it was much easier and faster to define the image plane with the smallest orifice area when 3-D echo was used. CONCLUSIONS Transthoracic real-time volumetric 3-D echo provides accurate and highly reproducible measurements of mitral valve area and can easily be performed from an apical approach.

Feasibility analysis of a case-based reasoning system for automated detection of coronary heart disease from myocardial scintigrams

Myocardial perfusion scintigraphy is a noninvasive diagnostic method for the evaluation of patients with suspected or proven coronary artery disease (CAD). We utilized case-based reasoning (CBR) methods to develop the computer-based image interpretation system SCINA which automatically derives from a scintigraphic image data set an assessment concerning the presence of CAD. We compiled a case library of 100 patients who underwent both perfusion scintigraphy and coronary angiography to document or exclude the presence of CAD. The angiographic diagnosis of the retrieved nearest neighbor match of a scintigraphic input case was selected as the CBR diagnosis. We examined the effects of input data granularity, case indexing, similarity metric, and adaptation on the diagnostic accuracy of the CBR application SCINA. For the final prototype, sensitivity and specificity for detection of coronary heart disease were 98% and 70% suggesting that CBR systems may achieve a diagnostic accuracy that appears feasible for clinical use.

Artificial neural networks and spatial temporal contour linking for automated endocardial contour detection on echocardiograms: a novel approach to determine left ventricular contractile function

This study investigated the use of artificial neural networks (ANN) for image segmentation and spatial temporal contour linking for the detection of endocardial contours on echocardiographic images. Using a backpropagation network, the system was trained with 279 sample regions obtained from eight training images to segment images into either tissue or blood pool region. The ANN system was then applied to parasternal short axis images of 38 patients. Spatial temporal contour linking was performed on the segmented images to extract endocardial boarders. Left ventricular areas (end-systolic and end-diastolic) determined with the automated system were calculated and compared to results obtained by manual contour tracing performed by two independent investigators. In addition, ejection fractions (EF) were derived using the area-length method and compared with radionuclide ventriculography. Image quality was classified as good in 12 (32%), moderate in 13 (34%) and poor in 13 (34%) patients. The ANN system provided estimates of end-diastolic and end-systolic areas in 36 (89%) of echocardiograms, which correlated well with those obtained by manual tracing (R = 0.99, SEE = 1.44). A good agreement was also found for the comparison of EF between the ANN system and Tc-radionuclide ventriculography (RNV, R = 0.93, SEE = 6.36). The ANN system also performed well in the subset of patients with poor image quality. Endocardial contour detection using artificial neural networks and spatial temporal contour linking allows accurate calculations of ventricular areas from transthoracic echocardiograms and performs well even in images with poor quality. This system could greatly enhance the feasibility, accuracy and reproducibility of calculating cardiac areas to derive left ventricular volumes and ejection fractions.

Paper: On using feedforward neural networks for clinical diagnostic tasks

In this paper we present an extensive comparison between several feedforward neural network types in the context of a clinical diagnostic task, namely the detection of coronary artery disease (CAD) using planar thallium-201 dipyridamole stress-redistribution scintigrams. We introduce results from well-known (e.g. multilayer perceptrons or MLPs, and radial basis function networks or RBFNs) as well as novel neural network techniques (e.g. conic section function networks) which demonstrate promising new routes for future applications of neural networks in medicine, and elsewhere. In particular we show that initializations of MLPs and conic section function networks--which can learn to behave more like an MLP or more like an RBFN--can lead to much improved results in rather difficult diagnostic tasks.

Assessment of coronary stenoses by Doppler wires: a validation study using in vitro modeling and computer simulations

The present study evaluates the use of intracoronary velocity measurements by Doppler guidewires for assessing coronary obstructions. In vitro experiments were performed in a flow model using acrylic phantoms of coronary stenoses with different configurations (stenosis area: 56%, 75% and 89%; stenosis length: 1 and 5 mm; stenosis border: tapering or abrupt). Nonpulsatile laminar flow conditions of a test fluid were established at flow rates ranging from 0.5 to 2.0 mL/s to simulate baseline flow and flow after vasodilation. Peak Doppler velocity was measured proximal to, within and distal to the model stenoses. Computer simulations were employed to calculate radial flow profiles with and without a Doppler wire aligned with the vessel center. In 84 in vitro flow experiments, peak Doppler velocity correlated well with the average flow velocity as calculated from the actual flow rate and the vessels cross-sectional area proximal to (r = 0.98, SEE = 1.4, p < 0.001) and within (r = 0.97, SEE = 16.4, p < 0.001) the stenosis. However, the ratio of calculated average velocity to Doppler-measured peak velocity was significantly different from 0.5, the expected value for a parabolic flow profile (0.76+/-0.08, 0.81+/-0.14; p < 0.001). Acceptable accuracy was found for the Doppler estimation of stenosis severity using the continuity equation (error: 0.9+/-1.2% and -4.6+/-3.5% for stenosis with a length of 5 mm and 1 mm, respectively). Doppler velocity reserve significantly underestimated the true flow reserve for the 56% and 75% stenoses (p < 0.01). Computer simulations demonstrated significant alterations of flow profiles by the wire, which explained the observed underestimation of the true flow reserve by the Doppler velocity reserve. Thus, Doppler guidewire measurements of intracoronary flow velocities are useful to assess the severity of coronary stenoses. However, the in vitro results and computer simulations indicate that guidewires alter the flow profile, so that Doppler velocity reserve may underestimate the true flow reserve.

Assessment of left ventricular function: comparison between radionuclide angiography and semiquantitative two-dimensional echocardiographic analysis

Measurement of global left ventricular function is important in the follow-up of cardiac patients and is a good prognostic indicator in acute cardiac situations. We compared quantitative measurements of global left ventricular function made with radionuclide angiography (RNA) and contrast cardiac ventriculography (CVG) to visual semiquantitative estimates from two-dimensional echocardiographic images (2D-echo). Three hundred and thirty-nine consecutive patients who underwent RNA were assessed with 2D-echo within 3 months. In addition, 92 of these patients also underwent CVG (correlation of ejection fraction between CVG and RNA:r=0.82;P<0.0001). The RNA mean ejection fractions in the four 2D-echo groups (0=normal, 1=slightly, 2=moderate, or 3=severe reduced left ventricular function) differed markedly (P<0.0001); however, there was overlapping among the groups (2D-echo score/RNA ejection fraction: 0=57.3%±12.8%; 1=46.0%±12.9%; 2=29.6%±12.2%; and 3=24.6%±11.5%) and the difference between 2D-echo scores 2 and 3 was not significant. 2D-echo showed a good concordance in RNA classes (0=≥505; 1=35%–49%; 2=21%–34%; and 3=≤520% ejection fraction) 0 (133/166; 80%) and 3 (18/30; 60%) but low concordance in classes 1 (27/82; 33%) and 2 (21/61; 34%). For accurate assessment of global left ventricular ejection fraction, visual semiquantitative judgement of a 2D echocardiographic image is limited in comparison to CVG or RNA, especially in patients with a slight or moderate reduction in left ventricular ejection fraction.

Transesophageal versus intracoronary Doppler measurements for calculation of coronary flow reserve

OBJECTIVE The present study was performed to compare coronary flow reserve by transesophageal Doppler echocardiography and intracoronary Doppler flow wire measurements in patients with LAD disease. METHODS 17 patients with various degree of LAD stenosis were studied. Intracoronary LAD Doppler measurements were performed at baseline and after intracoronary injection of 18 micrograms adenosine. Transesophageal coronary sinus and LAD Doppler measurements were performed at baseline and after intravenous dipyridamole (0.6 mg/kg/5 min). Coronary flow reserve was calculated as the ratio of hyperemic to baseline average peak velocities. RESULTS Coronary flow reserve was 2.44 +/- 0.62 and 2.19 +/- 0.76 for proximal and distal intracoronary measurements and was 2.25 +/- 0.64 and 1.74 +/- 0.63 for transesophageal LAD- and coronary sinus measurements. Proximal intracoronary flow reserve significantly correlated with transesophageal coronary sinus (r = 0.73, p < or = 0.001) and LAD (r = 0.70, p < or = 0.005) measurements, whereas distal intracoronary flow reserve only correlated with transesophageal coronary sinus flow reserve (r = 0.56, p < or = 0.02). Receiver operating characteristic curve analysis demonstrated similar diagnostic accuracy of all applied techniques for detection of a significant LAD stenosis. CONCLUSIONS Coronary flow reserve by both transesophageal techniques correlated with intracoronary Doppler flow wire measurements, however considerable discrepancies may occur in the individual patient.

SCINA: a case-based reasoning system for the interpretation of myocardial perfusion scintigrams

The purpose of this study was to utilize case-based reasoning techniques of artificial intelligence to develop a computer-based image interpretation system which automatically derives a comprehensive assessment concerning the presence, location, and severity of coronary artery disease from SPECT myocardial perfusion scintigraphy. Image analysis software was developed on a Macintosh computer. Polar map analysis of the tracer distribution in the myocardium was used to derive an input data set for the case-based reasoner. A case library of 100 patients who had been submitted to coronary angiography and perfusion scintigraphy was complied by a retrospective data base search. Sensitivity and specificity for detection of CAD were 83% and 70%. Sensitivity and specificity for localization of disease were 70% and 60% for the left anterior descending, 75% and 65% for the left circumflex artery and 53% and 63% for the right coronary artery. These results indicate that the case-based reasoning method permits the development of a clinically useful automated interpretation system for myocardial perfusion scintigraphy.

Concept Support as a Method for Programming Neural Networks with Symbolic Knowledge

Neural networks are usually seen as obtaining all their knowledge through training on the basis of examples. In many AI applications appropriate for neural networks, however, symbolic knowledge does exist which describes a large number of cases relatively well, or at least contributes to partial solutions. From a practical point of view it appears to be a waste of resources to give up this knowledge altogether by training a network from scratch. This paper introduces a method for inserting symbolic knowledge into a neural network-called “concept support.” This method is non-intrusive in that it does not rely on immediately setting any internal variable, such as weights. Instead, knowledge is inserted through pre-training on concepts or rules believed to be essential for the task. Thus the knowledge actually accessible for the neural network remains distributed or -as it is called-subsymbolic. Results from a test application are reported which show considerable improvements in generalization.