Segmentation of cancerous and non-cancerous lesions from dynamic contrast-enhanced magnetic resonans imaging of the breast by using image processing techniques

Early diagnosis of the breast cancer is one of the most important facts in life. Today?s methods of diagnosis show that any missing lesion can cause big problems in the future. By the help of early diagnosis of breast cancer, the mortality decreased to half. Besides , it also helps to protect the breast and provides more quality to life. These reasons make the early diagnosis of breast cancer a significant issue [14] . The goal of program which is composed in this study is to help doctors during the diagnosis. The doctors use software that includes basic image processing algorithms to interpret MR images. However, these software do not have any contribution to the decision of the diagnosis. These programs provide doctors to define the selected region of lesion. After evaluation of the particles from graphical and shape data, doctors can make a decision. However, the decision depends on the doctor?s carefulness and experience. Moreover, according to some doctors there can also be a malignant tumor that can be missed. The probability of missing these kinds of lesions has been considered as a serious problem. In this study, a method of early diagnosis is developed as a solution to decrease these problems The purpose of this study is to develop an algorithm and software that separates the cancerous lesions (malignant) and non cancerous lesions by the investigation of contrast holding matter and dynamic MR images. During the first scanning of patient, it is stated that the fat, muscles and normal tissues in the breast have different reactions as good manner and bad manner lesions. The program that is developed in the application evaluates 11 or 15 images in every measurement package internally and processes the outcome image into the last image in every measurement package. In the file that runs the program, 40 outcome images are formed .According to the profile images which is involved in the observation series the lesions which are determined as bad manner are painted to red and the good manner ones are painted to blue. In medicine science, the kind of behaviours of tumors under effect of contrast-enhanced compound had been observed by the doctors. From the inclination which is the reflection of these behaviours, the developed program has been compared the comparison between the methods of classification, local graphical analyse, image enhancement and the developed program. However it was obvious that the results obtained with the original algorithm is better than the other methods had been occurred. The images are provided from Sonomed Medical Image Center and Laboratory. During all stages of the studies performed on these images, it has been worked together with radiology specialists from Sonomed. The reports and existing pathological results of patients have been evaluated by the radiologists . Besides, radiology specialists approved the images, which are provided from developed program. To improve the study lots of ideas were shared. C programming language is used to develop this individual algorithm and it is planned to develop a user friendly interface within the earlier possible time. Key words: Breast cancer, MR, early diagnosis, image processing.