Often it is useful to make segmentation of anatomical structures in medical images. This facilitates qualitative representation of these structures, and measuring them.
Here we will explore some of the techniques that could be used for these purposes. We'll see preprocessing techniques and classification tissues of real clinical images. The classification result will be compared with the result that would give an expert
Given set of images X-ray 16bit from the medical database JSRT ([url removed, login to view]). The basic tissue structures that can distinguish an expert are the lungs, heart and bones (hand, clavicle, pelvis).
1.) In the training set of images try to find empirical thresholds so that you can isolate each of these areas to the extent [url removed, login to view] in a color image your results (different color to each type of area).
2.)You are given the ground truth for each of the categories of heart, collarbone, lungs. Build a histogram for each area using images training. You should be able to change the number of bins and see what changes.
3.)Try to locate the corresponding tissue structures (one of the 3 categories) in test images. For this purpose you will use the histogram as a measure of possibility to the pixel belongs in this category. We will define the number of bins(our job to define that)
a)normalize the histogram so that the sum of the elements of being one.
b)For each pixel of the image we will exam at what bin belongs-ranks and what value the bin has in each category. the category for which the corresponding bin has higher value wins the pixel.
c)If all bins are empty,the pixel does not belong anywhere and not colored.
d)Display with color your results for each image of the test images. Based on the ground truth give a table with the measure of the probability that the pixels of class i is classified as pixels of class j (i, j = 1,2,3).
4)Do the same by adding to training - test images
a)Noise gauss with mean 0 and p = 0.5,
b) Noise salt n pepper 5%
c) Use appropriate filter to refine your results.
I need comments in matlab so that i can understand what you do in each line