Cardiac wall motion detection by neural network analysis on Tc-99m MIBI myocardial perfusion gated single photon emission computed tomography

In this study we investigated cardiac wall motion detection by back propagation neural network analysis on Tc-99m methoxyisobutylisonitrile (MIBI) myocardial perfusion gated single photon emission computed tomography (GSPECT). Serial doughnut-like phantom images were generated to simulate cardiac wall motion of left ventricle, and as training data. After the neural network was trained, the neural network can perform motion detection for all serial phantom images and also for all series of patients´ GSPECT including short axis view, vertical long axis view, and horizontal long axis view images. The results of motion detection were displayed in the format of vector fields superimposed on the original GSPECT. The study showed that cardiac wall motion detection by back propagation neural network was useful in the evaluation of Tc-99m MIBI myocardial perfusion GSPECT.