Biomedical subjects
Publications and source records attributed to K J Goldman.
A technique of interactively combining digital images using virtual color tables is presented. The general combination algorithm, a particular implementation of this algorithm, and an interactive approach to combining functional images are described. Combined phase and amplitude images derived from gated blood-pool studies are presented to illustrate the method, and other applications of the technique in nuclear medicine are suggested.
The authors describe a digital filter that greatly enhances the quality of gated cardiac blood-pool images. Spatial filtering is accomplished with a minimum-mean-square-error (Wiener) filter incorporating measured camera blur and Poisson noise statistics. A low-pass temporal filter is then applied to each pixel, with the cutoff frequency determined from measurements of frequency spectra in 20 patients. This filter was evaluated in routine clinical use for nearly one year and found to significantly improve chamber definition, delineate wall motion abnormalities better, and reduce noise. To quantitatively assess the effect of the filter on image interpretation, four experienced observers evaluated wall motion in a series of mathematically simulated left ventricular images. ROC analysis revealed that accuracy in assessing wall motion was significantly greater with the filtered images.
A versatile mathematical simulation of a beating left ventricle, as seen in a gated cardiac blood-pool study, is described. The basic model consists of a fifth-degree polynomial rotated about its long axis. Motion is mimicked by sinusoidally varying the lengths of the long and short axes. The quality of clinical images is simulated through addition of Poisson noise and camera blur, determined from the measured line-spread function for Tc-99m. To achieve added realism, the modeled ventricle can be inserted in place of the patient's actual ventricle in clinical images. The dimensions, activity, and ejection fraction can all be adjusted. In addition, regional wall-motion abnormalities can be generated in any of the walls. To illustrate applications of the model in testing and evaluating cardiac computer programs, the performances of two ejection-fraction programs are compared, and use of the model in evaluating digital filtering algorithms is discussed.
Separation of systolic and diastolic parameters in gated cardiac blood-pool imaging (RVG) was achieved with the retention of two harmonics in the Fourier-series representation of the time-activity curve. Regional and global analysis of left-ventricular peak filling rate (PFR) and time to peak filling (TPF) was performed in 18 control subjects, 20 patients with coronary artery disease (CAD) but with normal RVG (normal regional wall motion and ejection fraction, and 16 CAD patients with abnormal RVG. In regional analysis of CAD patients, the standard deviation of the TPF histogram identified 13/20 (65%) of normal RVG patients and 12/16 (75%) of abnormal RVG patients as abnormal. In global analysis of CAD patients, PFR values identified 10/20 (50%) of normal RVG patients and 11/16 (69%) of abnormal RVG patients as abnormal. Thus, left-ventricular systolic and diastolic parameters can be separately measured with retention of higher-order harmonics in the Fourier transform, and regional inhomogeneity of diastolic filling can be detected in CAD patients with normal resting ejection fraction and wall motion.