Radionuclide Single Photon Emission Computed Tomography Lung Ventilation / Perfusion Image Processing And Registration

Along with the development of equipment in nuclear physic domain day by day, image processing becomes one of the most important parts in nuclear medical physic gradually. The quantifying of registration of the SPECT radionuclide ventilation and perfusion lung scan image is one of the effective ways for diagnosing pulmonary embolism. In recent years, the quantitative analysis on ventilation and perfusion has progressed at a rapid rate. It validates and supplements the qualitative method which is often used in clinical. Regardless of the traditional diagnosis or in quantitative analysis, whether ventilation and perfusion image can match accurate or not is a prerequisite for the true correct diagnosis conclusion.It is necessary to making use of some kinds of applicable image processing technology in order to enhance the radionuclide lung scan images before the two matched and divided, for there are noise and radiation attenuation in the course of imaging.This paper proposes a method of artifacts identification according to the average grey value through row and column of the SPECT lung imaging separately, which meanwhile locates a rectangular region of interest including the lung. Then a fast two-dimensional Maximizing Entropic threshold segmentation method is utilized to distinguish the pulmonary lobe from the background. The contour of the lung is obtained by use of canny operator. Based on these former works, image registration by corner detection and Maximization of Mutual Information algorithm is automatically realized on the basis of processing the lung images of 16 patients with normal scans and 10 with PE.