How to train svm for binary classification? (Tooth Disease Detection)

I'm currently working on a project that involves using SVM to classify images whether if it has a disease or not. From the original images, I have extracted the regions of interest (those with diseases) and now I'm stuck on how to use those pixel values to train the SVM classifier. How can I create a dataset that can be used for training using those pixel values from the images (and also that can be used in the matlab classification learning app)? Thank you for your help!

[Image Analyst]

What are the two features that you believe distinguish between healthy and diseased? Clearly it's not simply the gray level distributions alone, though that may be one of the features. What do you think it is? What features to you think you should measure? Were you going to use traditional SVM or multiclass SVM?

[Patric Navarro]

Thank you very much for helping! I'm gonna use features such as intensity gradient saturation of the image. I highlighted the regions of interest and acquired its correspoding intensity gradient and saturation which is in a form of matrix. What should i do next? I have watch a lot of tutorials and documentation but I still can't figure it out. Thank you for helping!

[Patric Navarro]

And also I'm planning to use traditional svm

[Image Analyst]

Can you take your two feature vectors and put them into classification learner? Also attach them here if you want.

[Patric Navarro]

I'm also not sure whether my feature vector is correct, what I did was after acquiring the pixel values from the ROI, i converted them into a series of column matrix where one column represents one image. Now, each column has 500 rows, is this correct or is 500 rows per column too much ? Thank you very much

[Image Analyst]

So one feature is basically just a collection of pixel values in your ROI. OK, but usually a feature will be that collection boiled down some, like to the histogram or the mean and standard deviation.

So that one feature is basically the intensity of your ROI, but very importantly, what is your other feature? You need at least two features to do SVM. Like I said, I know that intensity or even intensity distribution will not be enough to find healthy or diseased regions. You will need some other metric. What is it? Have you discussed this with your dentist or project sponsor?