See the article 'Image segmentation in digital mammography: comparison of local thresholding and region growing algorithms' (https://www.ncbi.nlm.nih.gov/pubmed/1394079).

See also 'Similar articles' on the right side of mentioned Web page.

Dear Sir,

The Trainable Segmentation

The aim of the segmentation is to isolate the diasese from the rest of the image. The process starts by first selecting a certain number of images as training images. From each training image, a number of samples, i.e. a small window of square regions of a certain size (e.g. 3  3), are taken from the inside of the diases (Class 1) and outside the diases (Class 2).

thank you Alexander, I saw this paper but I want your opinion for the best methods used in this case.

Brother Mazin,

Can you guide me to obtain the Trainable Segmentation algorithm in matlab.

See also 'Similar articles' on the right side of mentioned Web page.

thanks Alexander.