As explained above, the HRIMAC project was developed as a CAD system to assess the diagnosis of breast cancer. However, in this previous work, no algorithm for mass segmentation was proposed neither studied. This was a serious drawback to make a reliable tool for the radiologists. Thus,
However, it should be clear that our goal does not include the characterization nor the diagnosis of the masses found.
We split the main goal of the thesis into a set of sub-objectives. In this sense, the first sub-goal is a qualitative and quantitative study of the different proposals for mass segmentation. From this review, we notice that there is not a single method providing the best segmentation results in all cases. Moreover, we demonstrate that the algorithms depend on the density of the breast, and also, on the shape and size of the mass.
Hence, we want to design a new mass detection tool which takes the above three parameters into account. Note that, while the density of the breast is an information which can be known a priori, the shape and size of the mass is an information known a posteriori, once the mass has been found.
We divide the construction of the detection algorithm into two different steps: firstly we will design the algorithm without the breast density information, and subsequently we will introduce this information. Thus, the second sub-goal is the development of a new algorithm for mass detection which introduces mass shape and size information.
The result of this algorithm will be a set of regions, some of them being really mass and others depicting only normal tissue (a false positive). In this sense, the third sub-goal is the design of an algorithm for false positive reduction in order to increase detection accuracy.
The following step is the introduction of the breast density into the proposed algorithms. Thus, the fourth sub-goal is the construction of an algorithm capable of effectively classifying the breast according to its internal density before the application of the mass detection algorithm.
Finally, the fifth and last sub-goal is the introduction of the computed breast density information into both the mass detection and the false positive reduction algorithms.