I want to remove the extra boundaries from liver in CT images after the initial segmentation. how to do this?

Hello everyone. I want to remove the extra boundaries from the liver in CT abdominal images dataset. I segmented the liver from CT image but the boundaries of the liver also miss classified during my deep learning training. I want to remove it. and also may be another organ which miss classified. In the the following two images red line is the correct liver and around the liver having a boundaries which were miss classified and wrongly segmented. In green line the other organs also wrongly segmented. If anyone know this how to remove or refer any post processing method so please help.

image2

[Manjunath R V]

can u share the code related to the liver segmentation even i want the code for liver segmentation

[Image Analyst]

See this link: http://www.visionbib.com/bibliography/medical867l1.html#Liver%20Disease,%20Tomography,%20CAT%20Analysis

[ashutosh singh]

can u provide the code for liver segmentation from ct images

[Image Analyst]

@ashutosh singh, like I told @Manjunath R V, I don't have it. You'll have to write it yourself after finding a paper in the link I gave above that shows you how to do it.

Answer 1

[Image Analyst]

Turn the red and green coordinates into a mask, then mask them out. Full demo attached. Basically

% Make mask from red outline:
mask = poly2mask(redX, redY, rows, columns);
% Enlarge mask
mask = imdilate(mask, true(41));  % 41 is the size - enlarge or reduce to make the mask the size you want.
% Mask the image
grayImage(mask) = 0; % Zero out inside the mask.

[Ahmad]

Thanks image analyst its work. You are great. but if user don't select the area is it possible?

[Image Analyst]

Of course. You can create the mask in anyway you want, for example by thresholding.