How to consider each lesion area as a separate sample and extract features for training?

I have fewer number of images for my project but every image has more than one lesion area so after segmentation, I want to consider every lesion area as a separate sample for the feature extraction step in order to increase the number of training samples. I got the idea from this paper 'Ra, S., Suhilb, M., & Guruc, D. S. (2015). Segmentation and classification of skin lesions for disease diagnosis. Procedia Computer Science, 45, 76-85.' but can not apply it in code.

A sample image after segmentation.

Any help is appreciated.

[Ganesh Regoti]

Hi,

The question looks a bit vague. Can you explain what all you have tried so far? May be that could help us in understanding the problem.

[Mohammad Farhad Aryan]

Thank you Ganesh Regoti for considering my problem.

I am working on a research project about skin disease recognition. I don't have enough images for training svm algorithm so I want to make my training samples more from availabe images by taking each lesion as a separate sample and extracting features using glcm for each lesion. For example if I have 10 images and 5 lesion per image and consider each lesion as a separate sample then I will have 50 samples that I can create glcm for each sample and extract features using the created glcms.

Some papers did the same way using block processing and some other techniques like the paper attached above but I am not clear how to implement it in Matlab.

Answer 1

[Ganesh Regoti]

Hi,

As per my understanding you want to implement given paper in MATLAB. Here is my answer

1. There is a paper referenced to identify the lesions in an image, may be some algorithm is used to identify it. You can make use of regionprops for that.
2. To make multiple samples after finding the lesions, you can use imcrop and save it as seperate file using imwrite.
3. For understanding on how segmentation is done in MATLAB, you can refer the following link.https://www.mathworks.com/help/images/examples.html?category=image-segmentation

Hope this helps!

[Mohammad Farhad Aryan]

Thank you very much Ganesh Regoti!