Retrieval Technology Based On Multi-view Case Of Mammographic Masses

With the continuous development of medical standards and medical facilities, breast cancer is still one of the major killer diseases to the physical and mental health of women. Computer-aided detection system of breast disease can more accurately detects the suspected lesion region of masses for helping doctors to diagnose the disease. It reduces the rate of misdiagnosis of breast cancer disease. The traditional single-view retrieval system is unable to meet the needs of society because of the limitations inherent. It has come into being the retrieval based on multi-view case studies of mammographic masses.For false positives and regional matching of suspicious lesions on two-view masses suspected lesion detecting, it establishes a double scale Sech templates matching scheme which is similar to the gray mass distribution. The use of contrast enhancement and morphological filtering for image preprocessing; Built coordinates system which based on the reference objects of breast nipples, chest wall and the center axis; Construct the corresponding bar area which would be detected with a match in medio lateral oblique view based on the regional center location distance to the central axis in cranio caudal view of mammographic mass lesion; The mammographic masses suspected lesion region detected by the method that the overlay based on the two scales Sech templates and the result detected in morphological processing. Then it composed matching mutual information of the two view mass region.In order to suppress the over-segmentation of traditional watershed algorithm presence and achieve accurate mammographic masses edge profile, it has improved the algorithm in the segmentation process on region of interest. The use of two-dimensional Gaussian filter factor can do the filtering for the mass region of interest image. The small areas are obtained by dividing the crudewatershed. And the algorithm will mark the descending order gray value of each sub-region. Then the same or similar to the mean gray sub-regions are going to be merged. Take operations of morphological opening before removing the false positive regions. It will achieve accurate image segmentation of mammographic masses. Mammographic masses characteristic of normalization and similarity measure is an important step to multi-view mammary case retrieval.Combined with the query image library feature database, the image database is to be queried, and this article has 100 experimental samples of mammographic images of different retrieval modes and different similarity measure single and multi-view for mammographic masses retrieval. Compared the results of single-view cranio-caudal retrieval, single-view medio lateral oblique retrieval,multi-view consistent retrieval and multi-view arbitrarily retrieval, the multi-view retrieval results have a higher accuracy.