Lung Nodules Detection Based On Logical Constraints

At present,lung cancer has become a top killer that threats to human health,but lung cancer is characterized by nodular form at the beginning,so the detection of lung nodules becomes very important.The main problem in computer aided detection is that a large number of false positive detection,reduce the efficiency of detection.To solve this problem,this article stars proximal support vector machine(PSVM),and focuses on logic constraint’s learning algorithm,main contributions are summarized up as follows:The first,design and implementation of a lung nodule detection algorithm based on PSVM.The region of interest(ROI)detection and feature extraction are performed on the image,Selection of roundness,gray value,compact,gray variance,and calculate the value of features,which are used as data in the lung nodule detection algorithm.PSVM’s algorithm is change inequality constraints of SVM into equality constraint,which solves quadratic programming problems and improves computational efficiency,so this article based on PSVM to complete lung nodule detection.In order to verify the diagnostic accuracy of this method,artificial samples were used to construct the training samples,and then the method was compared with the traditional SVM-based lung nodule detection algorithm.Secondly,lung nodule detection algorithm based on constraint logic is proposed.Mining prior knowledge of lung nodules,and based on the proposed first-order predicate,given first-order predicate logic expression,the inequality constraints of the logic constraint learning frame are given by the mathematical formula proposed in this article.Then the algorithm is used in the training sample to get lung nodule detection model for the lung nodule detection.The performance of the proposed algorithm is compared with that of the other three typical detection algorithms in this article.The results show that the proposed algorithm not only improves the diagnosis precision,but also can provide the causal explanation for the further accurate detection.At the last,the two detection algorithms mentioned above are combined with the other two classic detection algorithms to design and implement the lung nodule detection system.