DBT Image Reconstruction And High Density Object Artifact Elimination Via Projection Enhancement

Digital Breast Tomosynthesis(DBT)is a new imaging technique for breast disease screening,which reconstructs three-dimensional breast images from projections acquired with a limited angular range.Such modality can effectively solve tissue-superposition problem in mammographic imaging,and therefore provide more diagnostic information and reduce the chance of misdiagnosis.However,the low-absorption rate of the breast tissue and the limited-angle scanning mechanism pose two major challenges for DBT from obtaining high-quality images.First,the attenuation coefficients of X-rays in the breast tissues are nearly homogeneous,resulting in DBT images with very low tissue contrast.Second,a small number of projections are usually acquired within a small angular range,resulting in out-of-plane artifacts and undershooting artifacts around the high density objects in the reconstructed DBT image.In this thesis,two strategies are specifically designed to solve these two problems,which are described as follows:(1)A projection enhancement algorithm named BOL-CLAHE is proposed based on block overlap limiting contrast histogram.The BOL-CLAHE uses the local information of the projection image to enhance its contrast,and removes the blocky effect in the aggregation process through the improved image partitioning strategy.Moreover,the guided filtering is incorporated into the projection enhancement framework to reduce noise in the projection image.In addition,the dense glandular tissue is segmented out from the breast area for grayscale stretching before enhancement,further increasing the contrast of the dense tissue.To verify the effectiveness of BOL-CLAHE algorithm,the simulated breast phantom and clinical patient data ware adopted for the experiments.The results of quantitative and qualitative analysis show that the proposed algorithm can effectively enhance the contrast of DBT images.The edges and texture information of the image shows that the lesion area is clearer in the images reconstructed from enhanced projections.(2)A weighted back projection strategy based on semi-confidence interval filtering is proposed to reduce the out-of-plane artifact and the undershooting artifact.Specifically,for a given pixel in the image volume,the corresponding points in the projection domain are firstly extracted for each view.Then,the gray range,mean and variance of these projection points are calculated.The points with high value are replaced it with the mean value according to the semi-confidence interval.Finally,a weighting strategy is incorporated into the backprojection process to further suppress the out-of-plane artifacts.On the other hand,a segmentation based method is adopted to reduce the undershooting artifacts during the filtering process.The results of quantitative and qualitative analysis show that the proposed algorithm can effectively suppress the out-of-plane artifacts and undershooting artifacts formed by high-density object.