Title :
Automatic Roi Selection for First Transit Nuclear Cardiology
Author :
Gallagher, J.H. ; Fouad, F.M. ; Cook, S.A. ; MacIntyre, W.J.
Author_Institution :
Division of Radiology Cleveland Clinic Foundation Cleveland, Ohio 44106 and Department of Scientific Computing
Abstract :
A fully automated chest parsing algorithm has been developed to delineate eight major regions of the upper body after a bolus injection of radiopharmaceutical. The algorithm uses the time-to-max functional image and depends on only one size parameter. Regions are selected by a 4-connected region growing procedure which is table driven from values extablished by an analysis of moments of the time-to-max image and region separation by iterative cluster optimization with a Fisher´s linear discriminant. On a set of 88 scans (59 cardiac outputs with 11 in heart failure and 29 pediatric shunt analyses), the chest parsings algorithm gave a satisfactory choice for the lungs, 89%; the right heart, 93%; and the left heart, 76%.
Keywords :
Biomedical imaging; Cardiology; Clustering algorithms; Heart; Humans; Iterative algorithms; Lungs; Performance evaluation; Pixel; Radiology;
Journal_Title :
Nuclear Science, IEEE Transactions on
DOI :
10.1109/TNS.1980.4330879