Automatic detection of the left ventricular outline in angiographs using television signal processing techniques

Segmentation of brain CT images using the concept of region growing

A method is described for extracting and isolating cerebrospinal fluid and tissue areas of brain images obtained with computed tomography. The classification of the pixels into components is based on region growing and nearest neighbor principles. To aid the performance of this method, the algorithm utilizes a priori information on the anatomic composition of the brain, and reduces the 'cupping effect' in the CT image that is attributed to beam hardening artifacts. In order to avoid subjectivity, the performance of the algorithm was tested by superimposing five computer-simulated circular lesions on different areas of the original CT scans, 8 mm thick. These images were taken at different levels in the brain, thereby accommodating different anatomy as well as the apical artifact of CT scanning. In this exploratory investigation, the false negative error of segmentation for lesions having diameter of 20 pixels was found in the order of 25% at an estimated partial volume (PV) effect of 50% that decrease further to about 5% for a PV of 80%. At that point the false positive error becomes the dominant error in the analysis.

Automatic contour definition on left ventriculograms by image evidence and a multiple template-based model

An algorithm which utilizes digital image processing and pattern recognition methods for automated definition of left ventricular (LV) contours is presented. Digital image processing and pattern recognition techniques are applied to digitally acquired radiographic images of the heart to extract the LV contours required for quantitative analysis of cardiac function. Knowledge of the image domain is invoked at each step of the algorithm to orient the data search and thereby the complexity of the solution. A knowledge-based image transformation, directional gradient search, expectations of object versus background location, least-cost path searches by dynamic programming, and a digital representation of possible versus impossible ventricular shape are exploited. The digital representation, composed of a set of characteristic templates, was created using contours obtained by manual tracing. The algorithm was tested by application of three sets of 25 images each. Test set one and two were used as training sets for creation of the model for contour correction. Model-based correction proved to be an effective technique, producing significant reduction of error in the final contours.

High precision quantitative angiography

Statistical considerations on the precision in the determination of blood vessel dimensions from digitized cine angiographic images are described. The resolution requirements related to "point measurements" and segmental diameter curve evaluations are discussed. The error associated with inaccurate determination of the vessel's centerline is analyzed. The concepts have been implemented on a high-speed image analyzing system, which measures blood vessel diameters with advanced automation. The performance of the system was evaluated with blood vessel phantoms, ranging in diameter from 0.88 to 6.26 mm. For these phantoms the minimum measurable change in vessel dimension over 20-pixel (-1.1 mm) long segments ranged from 3.4 to 0.2 percent, respectively.

Knowledge directed left ventricular boundary detection in equilibrium radionuclide angiocardiography

A method is presented that will automatically find the left ventricular (LV) boundary in any of several equilibrium radionuclide angiocardiographic (ERNA) image views. The lower level image analysis involves the use of directional gradient edge operators and an edge point linking scheme. The higher level portion of the algorithm uses a knowledge-based strategy that incorporates appropriate local and global information in heuristic cost functions to find complete boundaries. A detailed example of LV boundary delineation in the end diastolic frame from a left lateral view, ERNA image sequence is tracked through each stage of the proposed methodology. The use of this-approach for following the LV boundary through entire temporal sequences is also illustrated on 16 frames of both a left lateral view and a left anterior oblique view ERNA study.

Contour detection of the left ventricular cavity from angiographic images

A knowledge-based heuristic method is proposed to outline the left ventricular cavity from end-diastolic angiographic images in right anterior oblique projection in man. The algorithm search for the optimal path in a weighted graph, representing the state space, can implicitly be defined as the set of all the contour segments that can be produced by successor operators. Applications to routine images are presented.

Three-dimensional echocardiography for spatial visualization and volume calculation of cardiac structures

A new computerized image processing system was developed and applied clinically for three-dimensional visualization and volume calculation of cardiac structures, which were recorded in more than seven original two-dimensional echocardiograms in parallel planes. Three-dimensional display of this series of original two-dimensional echocardiograms was performed automatically using an overlaid display with different gray levels designating depths. The limitation of the present study was in the recording of original two-dimensional echocardiograms in parallel planes.

Automatic data processing in the cardiac catheterization laboratory

A review of automatic data processing in the cardiac catherization lab is presented. The emphasis placed on on-line manometry, indicator dilution and off-line quantitative ventriculography. The system organization is described and several specific examples given to illustrate the level of detail necessary to specify such a system. The clinical use of the system is described together with an evaluation of the original design goals in terms of the actual performance of the system. A few unexpected benefits have emerged with respect to quality control and reliability, but one of the original design goals, speeding up the catheterization procedure, has proven to be unrealistic. The next logical steps in the software and hardware evolution are described with the emphasis placed on making the system a more effective tool for the clinician.