Keywords:
Cardiac, SPECT, Computer Applications-3D, Computer Applications-General, Ischaemia / Infarction
Authors:
P.-M. Jodoin1, M. Havaei1, A. Oudot2, A. Lalande2; 1Sherbrooke/CA, 2Dijon/FR
Methods and Materials
Making the assumption that the LV has an anatomical U shape [Ale 10,
Ben 96],
3D sampling of the LV cavity is made in a spherical-cylindrical (S-C) coordinate system.
Conversion from Euclidean space to the S-C space is done by dividing the LV into a spherical part (the apex) and a cylindrical part (the base) (Fig.
1).
The only manual interactions are for the user to position the base and the apex of a T-bar shape as shown in Fig.2.
Then,
the anatomical U-shape of the LV is implicitly enforces because this step converts the U-shape into a quasi-planar surface in the S-C space (Fig.
3).
Then,
the myocardium centerline surface of the LV is located with a dynamic programming approach which,
in our case,
is performed with a graph-cut based energy minimization [Boy 01,
Boy 04,
Kol 04].
Finally,
epicardial and endocardial boundaries are outlined by considering the second statistical moment of the image.
Even if the LV has an attenuated signal due to a myocardial infarction,
the resulting shape will not contain any hole.
The method was tested on 18 rats.
Images were acquired with a small-animal γ-camera (Trifoil Imaging,
USA) after injection of a technetium sestamibi radiotracer (isotropic resolution = 0.7 mm per pixel,
12 cardiac phases).
The calculated cardiac cavity volumes and the ejection fraction were compared on the 18 examinations with the values obtained with a state-of-the-art and validated segmentation method (Flowquant software [Kle 10]).
Regression analysis was performed by calculated the correlation coefficient and the regression lines.