Abstract #3276

Automated Segmentation of Myocardial Infarcts on Delayed Enhancement MR Images

YingLi Lu¹, Graham A. Wright^1,2, Perry E. Radau¹

¹Imaging Research, Sunnybrook Health Science Centre, Toronto, ON, Canada; ²Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada

A novel, accurate and automatic algorithm is presented for segmenting myocardial infarcts from cardiac delayed enhancement MR images. The algorithm begins by localizing the LV center on a mid-ventricular slice by roundness metric. Then a bilateral filter is applied to smooth the image while preserving the edges. Next a tophat morphological transformation is used to isolate brighter objects. Finally a threshold is applied to determine the infarct. The correlation and Bland-Altman analysis of the infarct size derived from automatic and manual quantification indicated high correlations and good agreement.