![]() | Only 14 pages are availabe for public view |
Abstract Currently in clinical routine, it takes about 20 min by an expert to manually segment LV from stack of 2D short-axis CMR images (15) . Also, manual segmentation is prone to intra-and inter-variability. Therefore, as stated in section 1.3, designing an automatic technique to segment LV cavity from short-axis cardiac MR images was our aim and then use these segmented images to reliably estimate the LVEF with low user interaction. In order to achieve our aim, existing segmentation techniques were surveyed and limitations were spotted as shown in chapter 2. In this study, two limitations were selected: need for user interaction and sensitivity to ROI. In order to overcome these two limitations, the proposed approach used MSER detector (46) as detailed in chapter 3. To assess the accuracy of our approach, it was evaluated using 15 test sets provided during MICCAI 2009 LV segmentation challenge and results were discussed in chapter 4. The proposed technique proved its segmentation accuracy and robustness. |