Micro-calcification Clusters Detection Based On Non-sampled Shearlet Transform And PCNN Model

Breast cancer is a common malignancy.At present,breast cancer has become the major cause of cancer death in women.Because of the unknown cause of breast cancer,there is little measure can be adopted to prevent it.It has been proved that early diagnosis is an effective way to prevent the deterioration of the condition.Clinical trials show that nearly half of early breast cancers are diagnosed by the presence of microcalcification clusters(MCs)on mammograms,so the detection of MCs is of great importance in the early diagnosis and treatment of breast cancers.In this paper,a new method based on Non-subsampled Shearlet Transform(NSST)and Pulse Coupled Neural Network(PCNN)model is proposed to detect MCs in mammograms.The proposed method is roughly divided into three steps: image preprocessing,region of interest(ROI)extraction and MCs detection.In the image preprocessing stage,the method of maximum connected region is applied to remove the label and extra background,and a new algorithm based on top-hat transform and exponentiation operation is presented to enhance the breast region.In the ROI extraction stage,NSST transform is proposed for the first time to obtain MCs details.Finally,the PCNN model is used to detect MCs.In the experiment,lots of mammograms are used to test the algorithm,and the experimental results demonstrate that the algorithm presented in this paper is better than the other algorithms.Besides,the proposed method is simple and effective,and it can be considered to assist the radiologist for breast cancer diagnosis.The main work of this paper is as follows:1.A new enhancement algorithm based on top-hat transform and exponentiation operation is proposed to enhance breast region.It has been proved that top-hat transform has good effect on image enhancement.In this paper,we combine top-hat transform with exponentiation operation to enhance the breast region.Experimental results show that the algorithm can enhance MCs and inhibit the surrounding tissues and background,which lay a solid foundation for MCs detection.2.It is the first time to extract the region of interest of MCs using non-subsampled shearlet transform,and the results are compared with the results obtained by traditional wavelet transform and contourlet transform.The advantages and disadvantages of these three methods are introduced in detail.Through a large number of experiments,we found that NSST transform can not only detect the region of interest,but also highlight the MCs.3.In this paper,PCNN model is applied to detect MCs,and parameters setting of PCNN model is introduced in detail.In order to validate the effectiveness of the proposed algorithm,we use a large number of mammograms from MIAS database,JSMIT database and the People's Hospital of Gansu Province database to do the experiments.The results show that the MCs region detected by our algorithm are exactly the same as the region given by experts.By comparing with other MCs detection methods,the results show that our algorithm can get higher accuracy,specificity and sensitivity,and lower false positive rate,which can help doctors detect early breast cancer.