Research On False Positive Screening Of Thoracoabdominal Lymph Nodes In Three-Dimensional CT Images

As a very important immune organ in the body,the lymph nodes are spread all over the bodies and usually gather in groups.When the diameter of the short axis of the lymph node exceeds 10mm,it is considered to be a suspected metastatic cancer.The cancer research center in 2016,according to the survey data of lung cancer,stomach cancer,colon cancer,pancreatic cancer,liver cancer has become the fatality rate of the top ten diseases,and most of these cancers are distributed in the body’s chest and abdomen,usually by way of lymph node metastasis.Quantitative analysis of the progress of the thoracic and abdominal lymph nodes in the evaluation of certain diseases plays a vital role in accurate staging,prognosis,treatment selection and follow-up examination.With the rapid development of science and technology,the lymph node false positive screening of thoracic and abdominal CT images has become a research upsurge.However,the existing algorithms still have some problems:slow detection speed,high number of false positive nodes in the chest and abdomen,high computational cost and low sensitivity.In view of the above problems,this paper presents a method for false positive screening of thoracic and abdominal lymph nodes in three-dimensional CT images.Its main work is as follows:(1)We present an ROI extraction based on multiple views.The three-dimensional CT image data are decomposed into several two-dimensional angles of view,and the two-dimensional angle of view is optimized in each direction.The data of CT images are reduced,and the focus area can be observed and analyzed more comprehensively from different angles,thus speeding up the detection speed of the algorithm and improving the system performance.(2)A multiple feature extraction algorithm based on wavelet transform is proposed.The Wavelet Multiresolution method was introduced into the false positive screening of thoracic and abdominal lymph nodes,and different types of features were extracted in the wavelets domain.In addition to the high frequency part with a large amount of noise,the extracted features are better and more comprehensive,thus reducing the number of false positive lymph nodes in the chest and abdomen.(3)A PCA based feature combination optimization algorithm is proposed.The multi-group features were combined by permutation and combination,and the combination feature was optimized by PCA,and the optimal feature combination was selected by the convergence condition.It reduces the feature calculation,reduces the feature dimension,removes the redundancy and correlation of the features,thus reduces the computational cost of the algorithm and improves the detection sensitivity of the algorithm.In this paper,the FROC curve is evaluated by using the estimation method of 6 fold cross validation in the data sets of the open thoracic and abdominal lymph nodes.The experimental results show that the false positive screening method in three-dimensional CT images is effective,and compared with the existing methods,the computational cost is small,the real time is strong and the detection accuracy is high.