Application Of Artificial Neural Networks To CT Diagnosis Of Lung Cancer

Background and ObjectiveLung cancer is one of the most common cancer. Every year more than one million people die of lung cancer in the world. In developed countries, the mortality of lung cancer keeps the first place in cancer-related death. The incidence of lung cancer has risen to the second place in China, its mortality in the top of the urban population, and its incidence rate tends to rise. Lung cancer incidence in a short time, metastasis fast, prognosis is not satisfactory, and the overall five-year survival rate is only 15%. Early detection and early diagnosis can significantly increase the survival rate. Therefore, early detection, early diagnosis and early treatment are crucial to improve the survival rate of patients with lung cancer, the key to reducing mortality. However, due to a variety of related factors, early diagnosis makes a considerable limitation. How to improve the rate of early diagnosis of lung cancer is an urgent demand. Currently, there are three main ways in diagnosis of lung cancer: Imaging diagnosis,chemical diagnosis (serology and immunology) and cytology histology diagnosis. Imaging diagnosis includes X-ray imaging, CT, MRI, angiography and interventional radiology. CT is an important tool for the diagnosis of lung cancer which is one of the most widely used means of diagnosis. However, due to the complex nature of CT lung cancer and other uncertainty,it is not easy to make a correct diagnosis. If it is possible to create an artificial intelligence computer aided diagnosis system : Not only aid radiologists to avoid the subjectivity due to limited knowledge and experience,but also to distinguish features of lung cancer from other lung diseases on CT to discover lung cancer as early as possible, will greatly enhance the diagnostic value of lung CT. How to establish such a system is the key to solving the problem.Artificial neural networks (Artificial Neural Networks,ANN) is a nonlinear structural engineering information processing system which is on the basis of the simulation of the human brain mechanism of understanding and wisdom conducts. It is a cross-edge science in the integration of neural science, information science and computer science and had a rapid development in recent years. It is a simplification and simulation of information processing systems such as biological neural network structure function. Artificial neural networks has massive parallel processing, distributed information storage capacity, good adaptive ability, strong self-organization learning function and fault tolerance functions. At present, artificial neural networks have been widely used in finance, commerce, information, medical and other fields.In the field of medicine,Artificial neural networks has been used in clinical diagnosis of the disease (expert system), disease screening and diagnosis, disease-related factors studies,predict disease risk, survival analysis , gene identification, DNA and RNA sequence analysis and protein structure analysis. Artificial neural networks is also widely used in radiology, mainly in the imaging study of chest diseases. The application of artificial neural networks in chest imaging can improve the diagnostic accuracy of the diagnosis. Artificial neural networks has been successfully applied to the chest CT diagnosis abroad.The application of artificial neural networks to CT diagnosis of lung cancer has the potential to improve the diagnostic accuracy of lung cancer.The purpose of this study was to develop a computer-aided diagnostic scheme by using an artificial neural network (ANN) to assist radiologists in the distinction of benign and malignant pulmonary lesions and to evaluate the effect of artificial neural network output on the performance of radiologists using receiver operating characteristic analysis.Materials and Methods1. Lung CT 117 cases (58 cases of benign and malignant 59 cases), Matlab platform,SAS.2. Three experienced radiologists extracted 21 CT radiology features after they carefully observed the 117 CT cases.Then radiologists quantified the score of the 21 extracted radiology features. The 21 extracted radiology features combined with 5 clinical parameters constitutes the 26 input parameters of ANN schem. Then train the ANN to build models and blind test (simulation).3. Use trained artificial neural networks model to predict the forecast set. We compare the performance of artificial neural networks with logistic regression and experienced radiologists by means of ROC.Results1. Artificial neural networks with back-propagation algorithms, 6 hidden layer neurons, mc=0.95, achieve the desired goals through five iterations, the training stopped. The accuracy of the BP neural networks is 100%.2. Artificial neural networks and logistic regression forecast all samples of the training group and test group. The prediction accuracy of BP neural networks is 96.6% (113/117) and Logistic regression is 84.6% (99/117), the area under ROC curve of BP neural networks and Logistic regression are 0.986 (95%CI:0.944-0.998) and 0.909 (95%CI:0.842-0.954), respectively, with P =0.008.3. When forcast the test group, the diagnostic accuracy of BP neural networks is 90.9% (40/44) and the diagnostic accuracy of radiologists is 93.2% (41/44), the area under the ROC curve are 0.932 (95%CI:0.810-0.986) and 0.960 (95%CI:0.850-0.995) , respectively ,with P=0.568.Conclusions1. Artificial neural networks can be used in CT diagnosis of lung cancer as a potential useful tool.2. The performance of artificial neural networks is superior to Logistic regression.