I am doing texture analysis of Non contrast CT images. I intend to find out what texture feature is best to discriminate between lesions and normal tissues.
Dear Ohagwu
Hi
there are different effective texture features to discriminate between lesions and normal tissues. some of most important features are :
1. co-occurrence texture features of the wavelet approximation tumor region
2. gray level co-occurrence matrix .
3. textural based Haralick features (that is obtained from gray level co-occurence matrix).
I hope that they cloud be useful in your researches.
Dear Cristopher,
Actually, there are many you can use (see, for instance: http://www.cs.bris.ac.uk/publications/Papers/2000853.pdf).
Among them, I would start with LBP (code available here: http://www.cse.oulu.fi/CMV/Downloads/LBPMatlab), which proved effective in many applications, and related methods (http://dismac.dii.unipg.it/hep/).
Gabor filters could be another option - which however requires a bit more of tuning. There's a good implementation here: http://www.it.lut.fi/project/simplegabor/).
Hope it helps,
f.
- Badges
- Science topic