Image processing techniques in nuclear medicine

The application of image processing techniques to radionuclide images acquired on a gamma camera - computer system has been investigated. Hepatic perfusion imaging studies with 99Tcm-tin colloid were performed in patients with primary colorectal carcimma. The hepatic perfusion index performed poorly in the detection of those patients with occult or overt hepatic metastastes, as did mean transit times of liver colloid flow derived from deconvolution analysis. A discriminant function was developed which separated those patients with occult metastases from those without liver disease. A fully automatic algorithm to derive a left ventricular edge from each frame of an ECG gated cardiac blood pool study was developed and validated in patient studies. Left ventricular ejection fractions calculated from count rates within the edge were reproducible and correlated well with ejection fractions derived from the same images by a manual technique, and with ejection fractions derived from left ventricular cineangiography. Studies were performed in patients to evaluate the effectiveness of tomographic imaging of the myocardial perfusion imaging agent 99Tcm-tBIN for detection of ischaemic heart disease. Tomographic reconstructions in the planes of the axes of the left ventricle gave better results than transaxial reconstructions or planar imaging. Choice of the optimum reconstruction filter for use in tomography using 99Tcm-tBIN was examined by means of patient am phantom studies. These showed that more accurate diagnoses and better reconstructions were obtained with smoothing filters than by the use of sharp reconstruction filters. This work shows that optimum image processing techniques must be established to attain the best possible results with new radiopharmaceuticals for nuclear medicine investigations