Application of Artificial Neural Networks in Prediction Model of Early-stage Lung Cancer
Objective To establish prediction model of artificial neural networks through abstracting texture features of pulmonary nodules of CT images. Methods We abstracted texture features of pulmonary nodules of 2,171 CT images of 185 patients with gray-level co-occurrence matrix and wavelet transform from the outside diameter. Chi-square test and nonparametric statistics were used to analyze the database and we can get basic situation of cases and differences of texture features between benign and malignant cases. Finally, we established prediction model of artificial neural networks and evaluated its validity. Results Gender distribution in cases showed no statistically significant differences （=2.362, =0.124). We chose texture features which showed statistically significant differences 2 ? P between benign and malignant cases to establish prediction model of artificial neural networks. Sensitivity can reach 94.2%. Conclusions Establishing Prediction model of artificial neural networks based on texture features of pulmonary nodules of CT images can predict the characteristics of pulmonary nodules before pathological diagnosis. So it can aid to diagnose early-stage lung cancer.
Gray-level co-occurrence matriz Wavelet transform Tezture features Artificial neural networks Prediction model
Liu Yunning Wang Huan Guo Xiuhua Liang Zhigang He Qian
School of Public Health and Family Medicine, Capital Medical University, Beijing, P.R.China,100069 Department of Radiology, Xuan Wu Hospital, Capital Medical University, Beijing, P.R.China, 100053 Department of Radiology, Friendship Hospital, Capital Medical University, Beijing, P.R.China, 100050
国际会议
2008年国际应用统计学术研讨会(2008 International Institute of Applied Statistics Studies)
烟台
英文
1-5
2008-08-14（万方平台首次上网日期，不代表论文的发表时间）