Semiautomatic Detection

It has been our experience that the following characteristics are primarily used to differentiate stable from unstable caps using multicontrast MRI:

1. Appearance of a dark rim in 3D TOF images implying the presence of a thick cap.

2. Focal contour abnormality best observed in black blood images and implying a rupture or erosion of the fibrous cap.

The above two characteristics are used by the algorithm since they are the primary distinguishing characteristics. Absence of a dark rim is taken to indicate a thin or ruptured cap. Ruptured caps can be differentiated from thin caps by the presence of a focal contour abnormality. Other factors such as the presence of calcium near the lumen surface [98], flow abnormalities [7, 98], and intraplaque hemoharrage [97] may affect the correspondence between FC status and the dark rim but are not currently taken into account by the algorithm. To perform the FC evaluation, matched 3D TOF images and one black blood weighting are used by the algorithm to identify plaque status (Figure 8.25). Parameters for the dark rim are measured from the TOF image and those for focal contour

Figure 8.25: Schematic of algorithm—feature vector templates derived from reference datasets are used to classify points on the lumen contour based on their mahalanobis distance in the feature vector space.

abnormality are measured from one BB weighting. An operator draws lumen contours on both TOF and BB images by a Snake algorithm with appropriate weightings for the image energy term [100]. This step is the only semiautomatic step and the remaining steps do not require manual intervention.

In this approach, the algorithm classifies all points around the lumen contour, although the FC technically covers only the lipid core. With this approach, the human operator does not need to identify the body of the plaque and hence allows for easier automation. This is at most a minor limitation because normal wall is also associated with a dark rim. Hence, the algorithm classifies both types of stable walls, thick caps and normal vessel, as a single category. On the other hand, unstable fibrous caps that are thinned, eroded, or ruptured are separated by the algorithm.

0 0

Post a comment