Automated, robust and accurate image segmentation is an unsolved problem. I suggest you start by looking at some example works in this area. In our group, for example, we done some segmentation on spine and vessels, see:

spine:

Augmenting Auto-context with Global Geometric Features for Spinal Cord Segmentation: http://www.cs.sfu.ca/~hamarneh/ecopy/miccai_mlmi2013a.pdf

Globally Optimal Spinal Cord Segmentation using A Minimal Path in High Dimensions: http://www.cs.sfu.ca/~hamarneh/ecopy/isbi2013b.pdf

Spinal Crawlers: Deformable Organisms for Spinal Cord Segmentation and Analysis: http://www.cs.sfu.ca/~hamarneh/ecopy/miccai2006.pdf

vessels:

Vessel Crawlers: 3D Physically-based Deformable Organisms for Vasculature Segmentation and Analysis: http://www.cs.sfu.ca/~hamarneh/ecopy/cvpr2006.pdf

Live-Vessel: Extending Livewire for Simultaneous Extraction of Optimal Medial and Boundary Paths in Vascular Images: http://www.cs.sfu.ca/~hamarneh/ecopy/miccai2007a.pdf

Hi, Ghassan, thank you very much.

I will take a close look.

Because of their similar attenuation, for brain imaging the two techniques used are dual energy CT using bone removal based on differential attenuation of bone and contrast and bone subtraction based on a pre contrast scan CT.

Thanks Alex. will take a look.

BTW, I am now segmenting kidney and the associated tumor, so I basically focus on the spine and vessels around kidney.

If your goal is to isolate the kidney (and tumor) from surrounding structures of similar attenuation, then the most expedient approach may be to extrude manually drawn regions of interest augmented by region-growing to remove the spine and major blood vessels. While operator dependent and not particularly elegant, this approach will likely be the quickest way to remove the spine and enhancing blood vessels if you are dealing with up to a few hundred CT data sets. All commercial advanced processing solutions can do this, assuming that you have one available to you.

Ultimately, you need to articulate your specific goals in order to identify a satisfactory and efficient solution to analyzing your CT data.

Hi Geoffrey, Thank you.

What I am going to do is to obtain kidney, tumor, vessels and spines, which means to remove all the others from CT images.

Hi, Ghassan,

What kind of images (CT?) are you using?

Dear Ghassan,

I recently ran into this issue trying segment out the 6th rib for morphometric cortical analysis. We have recently adapted some the segmentation and analysis programs used in microCT for clinical indications. Specifically, we have modified the Skyscan software (Dataveiwer, CTAn, ect) to allow for the evaluation of the clinical CT data obtained on our Vereos dPET/CT and mobile Gemini CT systems. The programs allow us  to re-orient the VOI in the xy plane along the z-axis, segment and define ROIS/VOIs, perform morphometric 2D and 3D analysis, and perform 3D volumetric rendering. Once the data in the VOI is binarised the program will use edge detection and/or manual tracing to segment the tissue of interest. I am not sure if this will be helpful for your application, but after struggling with BoneJ we discovered that this was a major leap forward. 

Dear Fei, if you're still looking for a solution for this, please take a look at my company Simpleware's software for CT image processing at http://simpleware.com/, where free trials are also available.

Best,

Philippe