Methods for detection of left ventricular edges

The ability to accurately and reproducibly measure left ventricular performance offers significant clinical advantages in patient management. Specifically, data on wall motion of the left ventricle, the characteristics and shape of the left ventricular volume curve, and measurement of ejection fraction are the general parameters of interest evaluated. These parameters may be measured with either first-pass studies or gated equilibrium blood pool images. Either method is relatively simple, economical, and presents little risk to the patient. Over the last several years both methods have undergone considerable study, and relatively standardized techniques for the two methods exist at present. Both techniques require moderate to extensive data processing. In general, a region of interest (ROI) must be defined before further quantitative analysis is possible. There are at present multiple approaches to the establishment of an ROI for the left ventricle. The major differences between these approaches is in the algorithms used to generate the boundary of the ROI or "the edge". In order for the computer to recognize the edge of the left ventricle, objective and reproducible edge-detection processes are needed. It is the purpose of this paper to review computerized edge-detection algorithms as they apply particularly to the noisy and blurry images obtained in nuclear medicine studies.