| Medical imaging is one of the main means for cancer lesions screening. Due to the large number of images and the difference of filming equipment, it easily leads to visual fatigue and misdiagnosis for the doctors. As the "second pair of eyes", computer aided detection can help radiologists for disease diagnosis and decision-making. This thesis proposes a computer aided breast detection algorithm based on priori modeling. The main contents of the thesis are listed as follows:(1) To solve the problem of soft tissue organ segmentation, a segmentation algorithm is proposed based on a priori knowledge by this paper. Using the lasso model based on the edge priori knowledge information to initialize the target area. Obtaining an accurate soft tissue organ by an improved snake algorithm. The experimental results show that the proposed method can effectively segment the soft tissue organ from clearly CT images.(2) To overcome the problem of region of interest extraction based on the manual label or semi-automated manual label, a lesion region of interest extraction algorithm is proposed based on a statistical knowledge model. Our target is to build a statistical model for both negative and positive bags based on training samples. Finally, the average model and the unknown samples are calculated based on Euclidean distance. The biggest distance of samples is then regarded as the region of interest. The algorithm can fully automatically extract region of interest. The experimental results show that the proposed method has the higher accuracy for extracted region of interest.(3) To validly combine feature selection with the classifier, an improved algorithm based on LHs adaptive neural network fuzzy is proposed. The experimental results show that the proposed method can improve nearly2percentage points accuracy in lesion detection.In summary, the framework of the computer aided detection proposed in this thesis can effectively improve the performance of calcification clusters. |
PDF Full Text Request