Study Of Segmentation Of Suspicious Densities In Digital Mammograns

With the development of the medical science, people make more and more pressing demand for the early stage diagnosis of cancers. For instance, in the study of the breast' s cancer, people find that with the aid of CAD, the rate of the early stage diagnosis can be greatly raised. As the result, the patients will get more time for precautions and treatment and the incidence and death rate of this disease will drop considerably. The CAD software system (centering on pattern recognition) is widely entering the field of radiology, thus assist the doctor to make the diagnosis more precisely and more promptly.In the complicated algorithm system of CAD, the precise and effective Segmentation of suspicious densities in X-ray image (focus) is the key link. The result of the segmentation directly influences the final conclusion of the diagnosis. Accurateness and sensitivity of the segmentation immediately control TPF and FPF of CAD diagnosis. In the now available CAD system, there exist various algorithms for segmentation. Among them, the algorithms of pixel level (eg. threshold, region grow, expansion etc.) are mostly employed. These algorithms possess a rather high sensitivity, but on the other hand, they often make mistakes because they tend to be influenced by the noises and the normal tissue. In view of these issues, this essay advanced a two-step' s segmentation which is combined by pixel with region segmentation. In this method, through combining edge operator and ACM, attempting the segmentation of focus in mammogram, taking the points in contour which is acquired by edge operator as the original outline, viewing the GVF field in progress as the minimum energy of the external acting force, we take the acquired outline as the final one. Therefore, a higher sensitivity can be reached, the rate of segmentation of PFP can be reduced, and the false segmentation of the noises and the normal tissue decreased at the same time; In this way, the outline of the segmentation focus can be pressing more close to the actual tumor, and we can receive the accurate information about it.In the available study of CAD, most people only evaluate the result of the diagnosis, but not the segmentation. In the stage of segmentation, there is no unified method for the assessment of segmentation. It makes difficult for us to assess these segmenting methods, and to make contrast among them. This essay attempts to introduce phantom and ROC curve in the radiology into the assessment of segmenting algorithm. Thus, on the one hand the ROC parameter and TFP of the segmenting algorithm are received; on the other hand, a good deal of data and conclusions are got to support this algorithm.