A consultation expert system which supports our computer aided reporting system was developed. The system was used for the evaluation of the two dimensional polar (bull's eye) display of 201Tl myocardial SPECT. The system consists of patients management (PM) and consultation expert systems (ES). The former is connected to image processors coupled with scinticameras. The bull's eye display of myocardial SPECT is transferred from image processor to the data base of PM. When inference request is made, the feature extraction program extracts information on localization, extent and severity of focal defects comparing count-rates pixel by pixel with the reference obtained from seven normal controls. The inference engine is activated to determine presence of focal defects utilizing diagnostic rules in the knowledge base. The results are sent back to PM and reported with the probability of assurance. Fifty eight patients with old myocardial infarction (OMI), angina pectoris (AP) and other diseases as well as normal controls were included in the study. The decision for presence or absence of focal defects by ES agreed with that by nuclear physicians (NP) in 301 segments among 330 (91%) in stress images. The presence of redistribution in delayed images agreed in 43 segments among 67 (64%). Image interpretation by ES agreed well with that of NP in patients with OMI (19/20) and AP (9/11). Seven were interpreted as normal by both ES and NP. The system is useful, as it provides NP with complementary and supportive information applicable to decision making and reporting. Further clinical experiences can improve knowledge base for better ES function.

Computer Systems↗

PubMed1989

Primary spinal extradural hydatid cyst.

We report a rare case of spinal cord compression caused by a primary extradural hydatid cyst.

Adolescent↗

PubMed1987

[Development of a digitalized organ phantom for ECT simulation studies (DO-ECT PHANTOM)].

Explore the source record for details and available documents.

Models, Anatomic↗

PubMed1986

Clinical validation of fully-automated contour detection for gated radionuclide ventriculography with a slant-hole collimator.

A new fully-automated method for processing gated blood-pool images is presented and its clinical validity and performance for images with various noise levels are investigated using data obtained from a slant-hole collimator. The optimal preprocessing conditions are evaluated for the images with various noise levels to obtain the accurate ejection fraction, end-diastolic and end-systolic counts. This new method has successfully detected left ventricular contour in 92% of the 61 patients. The left ventricular ejection fraction obtained by the method related closely to that of contrast ventriculography (correlation coefficient r = 0.90). The end-diastolic volume also had a good correlation with contrast ventriculography (r = 0.90).

Filtration↗

PubMed1986

Automated body contour detection in SPECT: effects on quantitative studies.

To perform accurate in vivo quantitation by single photon emission computed tomography, we have developed a new method for detecting body contours for the correction of tissue attenuation. Our method can rapidly derive the best fit contours throughout the body by using the measured axial length of the patient body and Fourier filtering the detected contours which are defined by a unique bit-plane algorithm. We have also evaluated the effects of the body contours on the reconstructed images by using various attenuation correction techniques including a precorrection method (Sorenson, 1974), a postcorrection method (Chang, 1978), a weighted backprojection method, and a radial post correction method (Tanaka, 1983 and 1984). Counts in the specified region-of-interest in phantom images reconstructed by the radial postcorrection, weighted backprojection, and postcorrection methods were more strongly affected by inaccurately detected contours than were counts derived from images reconstructed by the precorrection method.

Body Surface Area↗

PubMed1986