Eye drop detection

[Shigool]

Hi

I'm trying to achieve eye drop detection using LabVIEW and Vision. My experience with NI Vision is limited and I'm trying to apply different filters. I was wondering if anyone in the community has tried to do this before? Does any one have experience with eye detection that could send me some example code please?

Kind Regards

[Bob_Schor]

What is "eye drop detection"?  Are you watching someone put drops in their eye?  You later mention "eye detection" -- are you trying to use a camera to detect an eye?  Human or animal?  What "question" are you asking (for example, do you want to know the orientation of the eye, and if so, in how many dimensions (six are possible)?

Bob Schor

[billko]

---

@Bob_Schor wrote:

What is "eye drop detection"?  Are you watching someone put drops in their eye?  You later mention "eye detection" -- are you trying to use a camera to detect an eye?  Human or animal?  What "question" are you asking (for example, do you want to know the orientation of the eye, and if so, in how many dimensions (six are possible)?

Bob Schor

---

Bob is our resident subject matter expert on this kind of stuff.  

[wiebe@CARYA]

Real-time face and eye detection in LabView using OpenCV Harr feature classifier - NI Community

LabVIEW Block Diagram of Eye Detection & Tracking Process  | Download Scientific Diagram (researchga...

Real-time eye tracking using a smart camera (archives-ouvertes.fr)

First results from bing...

[Shigool]

What I'm trying to achieve is to monitor a successful occlusion from an eye drop bottle to a human eye. I'm currently a research student and to cut a long story short I'm having to refactor my research.

The system I had in mind is to use a camera to take pictures before the dispensing of an eye drop and then after and judge or ascertain if the occlusion of the drop has been successful.

I'm pretty sure there has to be a way with NI Vision processing and I was toying around with some of the filters using eye pictures to try and detect differing locations of the eye but alas this is where my experience ends and I really could use with some help please! 

From my reading I've seen that a Gaussian and Canny filter can help, I'm going by the theory that there will be differences in the spectrum analysis or histogram of before and after eye drop dispensing.

Do you have any ideas?

[Bob_Schor]

Hello, Shigool.

     Please accept my apologies -- you really did mean "Eye drop detection", namely the detecting of an eye drop placed in the human eye.  The clue came from your response:

---

@Shigool wrote:

What I'm trying to achieve is to monitor a successful occlusion from an eye drop bottle to a human eye. I'm currently a research student and to cut a long story short I'm having to refactor my research.

---

Eye drop occlusion?  I asked a neurologist friend, who led me to articles on the Web about how to properly put eye drops in a patient's eye, and using "punctal occlusion" (pressure on the tear duct) to prevent the drug from getting into the rest of the body.

I've done (some) imaging of the eye, but for purposes of studying the eye as a motor system, namely its ability to rotate about multiple axes, and how to measure this (imaging is an interesting idea, but it is not so easy).

So here are some questions:

• Assume you have a stationary human eye (or a good model of the eye), one where motion is not an issue.  How would you use imaging to detect the presence and "distribution" of an eye drop?
• One possibility is for the drop to be colored, or fluorescent.  Is that what you are doing?
• If you are going to use an "awake, behaving" subject, how will you deal with blinks, eye movements, etc.?
• What sort of time course are you considering?  How will you prevent the drops from "leaking away"?

I have a colleague (from whom I learned about NI Vision) who studies lymphatic flow "in mice and men" using IR imaging.  I can tell you this is pretty challenging LabVIEW coding, one that could easily take a year or two of work, even with a colleague or two with whom to discuss ideas and problems.

Bob Schor

[wiebe@CARYA]

Vision can be very challenging. If you don't have much control over the environment (e.g. humans) it gets even harder.

First thing to do is to make reference images... Preferably under conditions controlled as good as possible (lightning (direction, color, intensity), camera distance, etc.). As many variations as possible. Also add expected results.

If you can't control image quality, you'd need even more reference images, as the scope of the problem is much larger.

Only then we can start to help.