Automated Sizing of Organoids in 2D

[NicholasW]

I am new to LabVIEW and I am attempting to automate the counting and sizing of organoids in NI Vision Assistant. I have looked through the guides but am having trouble figuring out where to start with this. The images are grayscale 2D images with both organoids and cells that failed to form organoids. The end goal would be the ability to input an image and having Vision Assistant output the number and size (SA or diameter is fine) of the organoids without counting the cells. Healthy organoids frequently manifest as a ring with a darker outline and lighter center. Once the basic functionality is running I will also aim to remove the counted organoids that display dead cells (usually via dark spots) to only count the healthy ones.

I have looked elsewhere online but haven't found any pre-existing methods for achieving what I am trying to do, however, if you know of one that I have missed I would appreciate a link to it. Aside from that I am really just looking for advice on what ways this problem could best be approached in Vision Assistant.

[altenbach]

To give us a better idea about the problem, can you post a typical picture. Also outline which parts you want you count.

[NicholasW]

I have attached an example image with the green outline representing what I want to count and red what I dont. 

[elcalverado]

 Hi Nicholas,

There's a couple different methods you could try, but it would take some effort of playing around with the function parameters to get it to work exactly how you want. I haven't tried out the methods I'll discuss below, so take this as a possible solution rather than a guaranteed one.

The first step is going to be to do some image processing in order to highlight the features in the image that you are interested in. Looking at the image you uploaded, it looks like the organoids on the surface appear pretty clear with good contrast, but the ones below the surface are blurrier and would be harder to distinguish from the surroundings. Are you looking to count every single one, even the cloudy ones?

Next you need to locate the organoids. If they are always pretty spherical, then using Shape Detection could work. You would set up the algorithm to locate circles in the image based on different parameters and thresholds. Once the circles are located, you could measure the organoids using the Clamp function. 

-Jordan