Finding Vessel tortuosity Of Fundus Images

I'm trying to find the vessel tortuosity of Fundus Images. I'm seperating the branches first. The problem is the code is not running as smoothly as it should be. Can someone please check the code.

name='Original';
fig=findall(0,'type','figure','name',name);
if isempty(fig)
    fig=figure('numbertitle','off','name',name);
else
    figure(fig)
end

% reading and displaying the image
I = imread('IM.jpg');
imshow(I);

grayImg=rgb2gray(I);
h(1)=imshow(grayImg);
% % imshow(grayImg);

eyeMask= im2bw(grayImg,graythresh(grayImg));
eyeMask= imfill(eyeMask,'holes');
eyeMask=bwareaopen(eyeMask,100);
% imshow(eyeMask);

%apply mask to eleminate background noise
grayImg(~eyeMask)=0;
set(h(1),'cdata',grayImg);

% Segment the vessels
vesselmask=edge(grayImg,'canny',0.10,1);
% imshow(vesselmask);
vs=imgca;

%%removing outer ring
vesselmask(~imerode(eyeMask,strel('disk',6)))=0;
% imshow(vesselmask);

% dilation
vesselmask=imdilate(vesselmask,strel('disk',5));
imshow(vesselmask);

*% vesselmask = bwmorph(vesselmask,'dilate');
vesselmask=bwmorph(vesselmask,'skel',Inf);
% vesselmask = bwmorph(vesselmask,'bridge');
% vesselmask=bwmorph(vesselmask,'skel',Inf);
vesselmask=bwmorph(vesselmask,'spur',5);
% 
% vesselmask=bwmorph(vesselmask,'skel',Inf);

imshow(vesselmask)*

branchPoints=bwmorph(vesselmask,'branch',1);
branchPoints=imdilate(branchPoints,strel('disk',2));
imshow(branchPoints)
bp=imgca;
% % 
% vesselmask = bwmorph(vesselmask,'dilate');
vesselmask=vesselmask & ~ branchPoints;
imshow(vesselmask)

[John D'Errico]

Repaired the formatting.

[new_bie]

Thanks for repairing the question. I'm new here. Tried a little bit to format it here. But did not know how to do it.

[Image Analyst]

I'm glad you're interested in learning how. Here are directions: http://www.mathworks.com/matlabcentral/answers/13205-tutorial-how-to-format-your-question-with-markup

Answer 1

[Image Analyst]

I didn't run it but just a quick scan looks like it should be okay. It should be splitting the vessel tree up into individual separate branches. What's the problem? I mean other than not being formatted correctly, which can be corrected if you read this.

You can just label the image and get the area with regionprops() to get the length of the branches, and get the PixelList and the endpoints from bwmorph() to find the distance between the endpoints. Then divide to get the tortuosity.

[new_bie]

I'm sorry. I'm new here and matlab is also new for me. Next time i'll try to format the question correctly.

When i applied this function on the vessel mask bwmorph(vesselmask,'skel',Inf) then there comes disconnectivity in all the branches. For removing it i again applied the dilation operation but then it could not find the branch points.

So its not splitting the vessel tree into exact branches. Please see the code again or check it by running it.

[Image Analyst]

No, it doesn't disconnect anything. Look at the image in the variable editor and you'll see the branches are intact. It may look like it's broken if it has to subsample a big image for display though.

What do you mean "again"? What are you removing "again"? When was something removed the first time?

You don't dilate then find branch points. You found branch points and then you dilated them - just look at your code again. I presume you did this to make the "hole" at the branchpoints a little bigger prior to erasing that area.

[new_bie]

oh Sorry!! i commented those dilation lines in the code

the last line vesselmask=vesselmask & ~ branchPoints is also branching out those lines which are not actually branches. it is splitting the long arcs into two arcs.

[new_bie]

here is the image

[Image Analyst]

Is that the same image as up toP? It doesn't look like it. Look at the original binary image, after you skeletonize but before you detect branchpoints. Is there a small branch going off of it? There probably is.

[new_bie]

I'm much obliged for your patience Image Analyst.

Yes the above image is the maximized version of the vesselmask after applying vesselmask=vesselmask & ~ branchPoints

the following figure (skeletonisedImage.fig) i got after skeletonizing the binary image and the other figure i got after applying this operation vesselmask=vesselmask & ~ branchPoints

Plus i could not find any branch going off of it.

[Image Analyst]

Sorry - it's not the same. The shape in branches.png does not show up anywhere in these figures. Just attach your original image and all the code necessary to run your program.

[new_bie]

The code and original image is attached.