Min Jin Lee, Presenter: Nothing to Disclose

Helen Hong PhD, Abstract Co-Author: Nothing to Disclose

Whal Lee MD, Abstract Co-Author: Nothing to Disclose

Eun-Ah Park MD, Abstract Co-Author: Nothing to Disclose

We have developed the automatic and time efficient segmentation method of left ventricle cavity and myocardium in MDCT images for the function analysis.

The quantitative analysis of left ventricle (LV) function such as ejection fraction and myocardial wall thickness in cardiac MDCT images is an essential for the determinant of pump failure after myocardial infarction and the diagnosis of heart disease. For the LV function analysis, endo-cardium and epi-cardium boundary extractions are a mandatory preprocessing step. However, it is difficult to extract them due to the heterogeneity of the myocardium density and lack of clear delineation between myocardium and adjacent anatomic structures. In this paper, we propose an automatic segmentation of left ventricle cavity and myocardium in cardiac MDCT images for the LV function analysis.

Our method has been applied to fourteen patients taken from traversing short-axis length from the base to the apex. The left ventricle cavity is segmented using adaptive thresholding and 3D seeded region growing. To involve papillary muscles in the cavity, the endo-cardium boundary is extracted by using ray-casting based intensity profile search. The epi-cardium boundary which has myocardium intensity and maximum gradient is extracted by using ray-casting based intensity profile search. To preserve a circular shape, the endo-cardium and epi-cardium boundaries are refined by using curvature-based spline fitting. The performance of proposed method (A) was evaluated by comparing with manually traced endo-cardium and epi-cardium boundaries (B,C) by two radiologists. The Dice coefficient was 0.92±0.02 for A vs. B and 0.93±0.02 for A vs. C and 0.95±0.02 for B vs. C, respectively (all p > .05 by paired t-tests). The whole process of our method was finished within 1 second.

Our endo-cardium boundary extraction consistently segments cavity including papillary muscles. Our epi-cardium boundary extraction delineates boundary regions without leakage to adjacent anatomic structures such as right ventricle and liver. In particular, our method robustly segments the myocardium boundaries in subject with low contrast density distribution.

Lee, M, Hong, H, Lee, W, Park, E, Automatic Myocardial Segmentation Using Multidetector CT for Assessment of Left Ventricular Function Analysis.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9014494.html