next up previous contents
Next: What are LSE results Up: How to segment in Previous: Semi-automatic segmentation   Contents


Segmentation principle

Image LSE_validroi In the LSE, the Delineation Tab is used to segment lesions on any of the loaded images. When a lesion is segmented (manually or semi-automaticaly), the obtained segmentation has to be validated on the correct image. First, select on which image the segmentation has to be applied (button on the smaller images bottom). Then press the validate roi button. See figure 6.14 for a step by step procedure description.

Figure 6.14: Step by step procedure for ROI validation.
Image LSE_roivalidation
The lesion has been segmented on the T2 image but I want to associate it with the PD image. 1) I select the PD image. 2) I press the validate roi button.
Image LSE_roionima
The segmentation is in the PD image tab.
 

Image LSE_roiresetIf you go back to the Delineation tab. You can reset the ROI on all the images of the Delineation tab thanks to the reset ROIs button. ROIs which have been previously validated are not reseted in the image tab.

Image LSE_roiaddYou can also realize a new segmentation that you can validated on another image or add to the PD image segmentation. Warning: If you validate the new segmentation on the PD image, you erase the former one.

Image LSE_roireload In the Delineation tab you can also reload a segmentation from an image tab thanks to the ROI reload button. The ROI validation, the ROI addition and the ROI reloading are realized on the image which is selected in the delineation tab.

Application: The region growing segmentation did not work perfectly on a lesion (see figure 6.15). Moreover, the obtained segmentation can also be reduce thanks to the manual segmentation (label 0 = eraser).

Figure 6.15: Region growing correction.
Image LSE_app1 Image LSE_app2
1) I clic in the middle of one lesion.
2) After a clic on the region growing button, I obtain a first unperfect segmentation of the lesion. I validate it on the T2 image.
   
Image LSE_app3 Image LSE_app4
3) I clic on a part of the lesion which was not correctly segmented. Then I clic on the region growing button. I add this new ROI to the T2 image.
4) I reload the ROI from the T2 image. The segmentation is acceptable.
   


next up previous contents
Next: What are LSE results Up: How to segment in Previous: Semi-automatic segmentation   Contents
Nicolas Toussaint 2007-06-22