What is the best way to detect white blood cells among the red blood cells in blood sample using opencv?

asked 2016-12-25 01:03:43 -0500

Pradeep0168 gravatar image

updated 2017-01-03 03:02:34 -0500

image description

image description

positive image negative image

image description image description

edit retag flag offensive close merge delete



what have you tried so far ?

please take a look at the faq , we do expect you to do some research on your own, before asking here.

berak gravatar imageberak ( 2016-12-25 01:47:40 -0500 )edit

I have tried using cascade classifier. I trained the classifier and then applied the classifier on test image but not satisfied with the result. Its giving me some wrong predicted reason of interest.

Pradeep0168 gravatar imagePradeep0168 ( 2017-01-02 00:19:00 -0500 )edit

Then please elaborate what you have been doing using the cascade classifier and where it goes wrong. Then we might actually be able to help you improve it.

StevenPuttemans gravatar imageStevenPuttemans ( 2017-01-02 06:03:08 -0500 )edit

For some of the test images it's not detecting properly.

Pradeep0168 gravatar imagePradeep0168 ( 2017-01-02 10:29:04 -0500 )edit

Read my post again...

StevenPuttemans gravatar imageStevenPuttemans ( 2017-01-02 10:34:42 -0500 )edit

The best technique to use depends on what kind of results do you need. Do you need to detect each cell individually? Do you need to realize if there are white blood cells present in the scene or not? Do you need to create some kind of ratio between red and white blood cells?

Pedro Batista gravatar imagePedro Batista ( 2017-01-02 11:52:48 -0500 )edit

The requirement is to detect all the white blood cell present in the image and demarcate the boundary of the white blood cell.

Pradeep0168 gravatar imagePradeep0168 ( 2017-01-02 18:23:11 -0500 )edit

Hi StevenPuttemans, the first small image i am using as a positive image and second small image i am using as a negative image.

Pradeep0168 gravatar imagePradeep0168 ( 2017-01-02 18:31:44 -0500 )edit

Ow why black out the cell in the positive image. This way you create artificial borders to the background, which is not there in your actual application and thus, your detector fails. Do not do that. You should read up on decent sample collection for machine learning models.

StevenPuttemans gravatar imageStevenPuttemans ( 2017-01-03 02:13:33 -0500 )edit

I have read up on sample collection. I have created the positive image by using above third small image and image mask which is fourth small image. Could you please tell which one should i choose positive and which one negative image to train? thanks.

Pradeep0168 gravatar imagePradeep0168 ( 2017-01-03 03:06:35 -0500 )edit