In this paper, we present a new approach for the extraction of brain tumor boundary in a series of 2D MR image slices. The shape and position of tumor in one slice could be assumed to be similar to that in its neighboring slices. Using this correlation between consecutive images, the initial plan applied for each slice is extracted from the resulting boundary of the previous slice. The tumor boundary is located using region and contour deformation, which tolerates a rough initial plan. Therefore, only one coarse manual initial plan is required for the whole series of MR image slices. Performance of our approach is evaluated on MR image set. Comparisons with manual tracing show the accuracy and effectiveness of our approach.
展开▼
