Abstract

Artificial intelligence and machine vision systems often use shape analysis as a fundamental approach to characterization of complex objects and motion. Such concepts have been used to advantage in industry but have not been extensively exploited for cardiac diagnosis.
Aims : (1) To develop a method of quantitative regional function analysis based on analysis of dynamics of ventricular shape that will replace traditional methods used to assess wall motion so that assumptions regarding coordinate, reference and indexing systems, geometry of the ventricle and uniformity of its contraction pattern can be abandoned. (2) To quantify the regional shape of the normal right and left ventricles and establish the patterns of shape abnormality seen in the setting of coronary artery disease. (3) To establish the relationship between regional shape and regional function. (4) To develop a clinically useable, automated method for ventricular feature extraction and determine its usefulness in predicting the presence and location of coronary disease by analysis of ventricular images obtained during an ischemic stress. (5) To assess the prognostic significance of regional shape analysis. (6) To automate the requisite edge detection by using integrative methods of artificial intelligence. The hypothesis is that the normal ventricle has a characteristic shape and rate of shape change throughout the phases of the cardiac cycle and that abnormal ventricles have abnormal shapes and shape changes by virtue of abnormalities of function. Therefore, quantitation of abnormalities of shape parameters will provide an indirect measure of ventricular dysfunction. This will be determined by frame-by-frame curvature analysis of cineventriculograms in patients with normal and stenotic coronary arteries. Animal studies will compare results of regional thickening and thinning determined by implanted sonomicrometers with shape changes assessed from simultaneous contrast ventriculograms and ventricular outlines determined from implanted subendocardial lead pellets. The success of this approach will allow abandonment of numerous controversial assumptions mandated by traditional approaches and will solve a long-standing problem in cardiology. The concepts to be explored will provide new and unique information of potential prognastic importance and it will advance the neglected field of computer vision in the study of non-rigid bodies in motion.

  Funding Agency

Agency
National Institute of Health (NIH)
Institute
National Heart, Lung, and Blood Institute (NHLBI)
Type
Research Project (R01)
Project #
1R01HL036813-01A3
Application #
3352103
Study Section
Diagnostic Radiology Study Section (RNM)
Project Start
1988-07-01
Project End
1991-06-30
Budget Start
1988-07-01
Budget End
1989-06-30
Support Year
1
Fiscal Year
1988
Total Cost
Indirect Cost

  Institution

Name
University of Michigan Ann Arbor
Department
Type
Schools of Medicine
DUNS #
791277940
City
Ann Arbor
State
MI
Country
United States
Zip Code
48109