Image Analysis of Cell Samples
in [Matlab]
|
From: Akshay Chaudhari on 24 Dec 2009 00:31 I am currently working on a project that involves image analysis on images of different cells (Not for a grade, just academic research). So basically I have images of a cells that have undergone different stains and I have to calculate the area of certain regions within the cell. So essentially I have a region of a cell enclosed by small periphery that is a complete different color. It is hard to describe, so as a reference, this link shows a sample image. http://www.ispub.com/ispub/ijpa/volume_7_number_2_15/synovial_hemangioma_of_the_elbow_an_uncommon_lesion_to_be_considered/hemangioma-fig4.jpg So essentially can anyone help me with this? I would really appreciate it a lot!
From: ImageAnalyst on 24 Dec 2009 10:16 Akshay Chaudhari It looks like you can just threshold, label, and regionprops. See my demo that walks you through this process: http://www.mathworks.com/matlabcentral/fileexchange/25157 For your image, you could either threshold on the light blobs, or threshold on the dark perimeters and then do a holefill. Then label (with bwconncomp or bwlabel) and call regionprops(). It's all copiously commented in the demo. Good luck, ImageAnalyst
From: Akshay Chaudhari on 30 Dec 2009 19:34 ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <b9e91f5a-f660-41df-af5a-fc84a88df244(a)u7g2000yqm.googlegroups.com>... > Akshay Chaudhari > It looks like you can just threshold, label, and regionprops. > See my demo that walks you through this process: > http://www.mathworks.com/matlabcentral/fileexchange/25157 > > For your image, you could either threshold on the light blobs, or > threshold on the dark perimeters and then do a holefill. Then label > (with bwconncomp or bwlabel) and call regionprops(). It's all > copiously commented in the demo. > Good luck, > ImageAnalyst Hey thanks a lot buddy, I really appreciate your help!! I looked at your # of posts and its really nice that you help others when you have the knowledge to do so. Great job and props. Have a great new year!
From: ImageAnalyst on 30 Dec 2009 19:56 You're welcome. My pleasure. You have a great new year too.
From: Dave Robinson on 3 Jan 2010 11:51
ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <f0bdf092-3abf-4321-a8b4-0ae2a6603e45(a)e37g2000yqn.googlegroups.com>... > You're welcome. My pleasure. You have a great new year too. Hi Whilst not having much to add to IA's excellent advice, one thing I noticed was that your images are very unsaturated, which means that there is a great deal of correlation between the colour planes - i.e if you look at the three colour planes as greyscale images, they are almost identical. One technique I have used to great effect on images of this category amounts to virtual cell staining - see the images I have put up here: - http://s931.photobucket.com/albums/ad152/dave_robinsonsnr/?action=postupload The first shows your original image, and the second shows my processed image. The technique amounts to finding a custom colour space that effectively decorrelates the three colour axes by undertaking a PCA process between the three colour planes. Once the images have been transformed into this space, you could then consider using a colour filtering technique to seperate your cells from the background, possibly simplifying your blob analysis suggested by IA. Currently I don't have a Matlab version of this process, but as soon as I have time, I will construct one, and place it in the File Exchange. Regards Dave Robinson |