Retina vessel segmentation

[Mr_Sonky]

Hi, I am new to ImageJ/Fiji and am also new to this forum. I'm currently working on a way to segment some blood vessels from a 16-bit grayscale image of a rodent retina. See image below:



I have tried all auto-threshold methods and manual thresholding but still no dice. The areas adjacent to the vessel have pixel values too similar to my vessel. Yet I can easily demarcate these vessels with my naked eye (manual segmentation)..

I've done a quick search of the forum but did not find anything useful... any help is appreciated!

[Olivier Burri]

Hi Mr_Skony,

It would make sense that no threshold methods works, as some vessels seem bright, and others dark. Similarily, the illumination is very uneven, so I highly doubt any kind of automatic segmentation would work in this case. How are these images acquired?

Also, perhaps you could provide a hand-drawn rendition of the structure you hope to extract as I am not sure what I am looking at and am by no means a rodent retina blood vessel expert :)

Best

Oli

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Dionysios Lefkaditis]

Hello Mr_Skony,

It sounds like you might benefit from a machine learning approach to
perform the required segmentation. I suggest that you should give
Advanced Weka Segmentation plugin a try:
http://fiji.sc/wiki/index.php/Advanced_Weka_Segmentation

cheers,
Dionysios

On Thu 24 Jul 2014 09:34:54 AM CEST, Burri Olivier wrote:

> Hi Mr_Skony,
>
> It would make sense that no threshold methods works, as some vessels seem bright, and others dark. Similarily, the illumination is very uneven, so I highly doubt any kind of automatic segmentation would work in this case. How are these images acquired?
>
> Also, perhaps you could provide a hand-drawn rendition of the structure you hope to extract as I am not sure what I am looking at and am by no means a rodent retina blood vessel expert :)
>
> Best
>
> Oli
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Ignacio Arganda-Carreras]

You may want to give Trainable Weka Segmentation a try:
http://fiji.sc/Trainable_Weka_Segmentation

I would start with just a few features and see how it goes.

Best,

ignacio


On Thu, Jul 24, 2014 at 9:34 AM, Burri Olivier <[hidden email]>
wrote:

> Hi Mr_Skony,
>
> It would make sense that no threshold methods works, as some vessels seem
> bright, and others dark. Similarily, the illumination is very uneven, so I
> highly doubt any kind of automatic segmentation would work in this case.
> How are these images acquired?
>
> Also, perhaps you could provide a hand-drawn rendition of the structure
> you hope to extract as I am not sure what I am looking at and am by no
> means a rodent retina blood vessel expert :)
>
> Best
>
> Oli
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
Ignacio Arganda-Carreras, Ph.D.
Seung's lab, 46-5065
Department of Brain and Cognitive Sciences
Massachusetts Institute of Technology
43 Vassar St.
Cambridge, MA 02139
USA

Phone: [phone]
Website: http://bioweb.cnb.csic.es/~iarganda/index_EN.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[James Ewing]

Your eye is looking at the edges and other relatively high-frequency (spatial) phenomena.  You might try a Fourier transform, filter the low-frequency components, and back-transform.  Alternatively, there are spatial filters that are basically high-pass filters - try a Gaussian high-pass filter.
  - Jim Ewing

On Jul 24, 2014, at 12:03 AM, Mr_Sonky <[hidden email]> wrote:

> Hi, I am new to ImageJ/Fiji and am also new to this forum. I'm currently
> working on a way to segment some blood vessels from a 16-bit grayscale image
> of a rodent retina. See image below:
>
> <http://imagej.1557.x6.nabble.com/file/n5008869/Eye_blood_vessel.png>
>
> I have tried all auto-threshold methods and manual thresholding but still no
> dice. The areas adjacent to the vessel have pixel values too similar to my
> vessel. Yet I can easily demarcate these vessels with my naked eye (manual
> segmentation)..
>
> I've done a quick search of the forum but did not find anything useful...
> any help is appreciated!
>
>
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Retina-vessel-segmentation-tp5008869.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Mr_Sonky]

Hi all!

It's been awhile since I had the chance to come back to this. Thanks Oli, Dionysios, Ignacio and James. I've given Weka segmentation and go and am really surprised how easy it was! Here is a sample of my handiwork:







Thanks!! I might have more questions later, but this seems to do the trick for now..