extraction of the segmented regions and feature extraction

[briannaburton]

Read the article A New Local Adaptive Mass Detection Algorithm in Mammograms.

Check other articles if needed. Define features for our masses. Check feature that Bria suggested: local binary patterns.

Output should be a function that will return a feature array for a given image. At the same time we should be able to extract

[OlehKSS]

@doiriel, please write your findings on Local Binary Patterns as a comment to this issue.

[OlehKSS]

A New Local Adaptive Mass Detection Algorithm in Mammograms.

Features that we can use for our classifier:

• perimeter of the region
• area of the region
• circularity = P^2 / 4piA, where P is circumference (perimeter of the region), A is area of the region.
• AC ratio = Area/Circularity, it is used to filter curve-linear structures such as blood vessels and milk ducts in mammogram.

An Efficient Approach for Automated Mass Segmentation and Classification in Mammograms

Features that we can use for our classifier: Besides geometrical features described in the previous article, they have used several features that imply operations with the image intensity as well.

• mean of intensity
• standard deviation
• smoothness
• pixel value fluctuation (not clear how to use it after all)
• conspicuity (not very clear how to do calculations on the backgounds)
• uniformity and a lot of others

In this article they used random forest classifier and claim it to be the best of the possible choices.