Automated quantification of rubidium-82 myocardial perfusion images using wavelet based approach

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT REQUEST OF AUTHOR.] The main work in this dissertation focuses on development of automated quantitation algorithms for Rb-82 PET images. The current methods for quantitation is model based and may fail to determine regional details as produced by higher resolution PET images. Thus automated approaches which are model-independent and detects details of the myocardium would seem valuable for accurate quantitation. Two major components of such an automated approach developed in this work are myocardial reorientation and myocardial boundary detection. The new reorientation algorithm is based on detection of myocardial contour with wavelet processing to detect the oblique angles of the myocardium. Evaluation of the reorientation algorithm is done for cardiac phantom and large number of clinical patients by comparing with user selected angles. The boundary detection algorithm is based on spherical sampling and analysis of counts based on gradient functions. For evaluation of the boundary detection algorithms, the results are compared with results from current techniques. For cardiac phantom and patients, higher degree of correlation and agreement is obtained for quantative parameters between current method and the new algorithm. Validation of the new algorithm with more patients and with results from high-resolution modalities will prove its robustness and may bring new advances in the quantitation of myocardial perfusion imaging.