how to extract features from images

The project is on BRAIN TUMOUR EXTRACTION USING MRI IMAGES using MATLAB and further determining whether the extracted tumour is MALIGN or BENIGN.

The project has been completed upto the point of brain tumour extraction.

The problem arising now is how to determine as to in which category this extracted tumour falls... MALIGN or BENIGN.

For this first of all various features of images need to be extracted depending upon which the tumour would be classified as malign or benign. I am currently using matlab 2011 version.

The MRI scanned images being used by me for this project can be seen at http://www.flickr.com/photos/77172895@N08/?saved=1

Answer 1

[Jorgen Kongsro]

Dear Tarun. I presume you have the reference material of the tumors which is MALIGN and which is BENIGN. If this is related to texture, I would try simple textural analysis like GLCM or more advanced AMT. If this is related to pixel value, I would try plotting the histograms of pixel values, and make a multivariate calibration of the histogram against your reference values.

Answer 2

[Image Analyst]

I don't know what features you should measure, if they are even detectable from MRI. The radiologist you are working with should know that and can tell you. If these are human images, you must have a radiologist and permission from the institution that captured the MRI scans, and probably the patients also, or else you wouldn't have any images due to privacy regulations.

I don't think you're pioneering new territory here so you can search the literature: http://iris.usc.edu/Vision-Notes/bibliography/contentsmedical.html#Medical%20Applications,%20CAT,%20MRI,%20Ultrasound,%20Heart%20Models,%20Brain%20Models for methods that people have developed, successfully used, and published. Pick one of those and try to code it up in MATLAB yourself, hire the authors as consultants, or buy the software. You can get help with MATLAB problems here.