文献类型：会议论文

英文题名：A Classification of benign and malignant lung nodules based on feature fusion and improved random forest

作者：Zhang, Wei[1]; Hua, Deliang[2]; Li, Mengting[3]; Wei, Lili[3]; Ren, Zhen[4]; Song, Jinxia[5]

第一作者：张伟;张维

机构：[1] School of Health Management, Gansu University of Chinese Medicine, The Collaborative Innovation Center for Prevention and Control by Chinese Medicine on Diseases Related Northwestern Environment and Nutrition, Gansu, Lanzhou, China; [2] Affiliated Hospital, Gansu University of Chinese Medicine, Gansu, Lanzhou, China; [3] School of Health Management, Gansu University of Chinese Medicine, Gansu, Lanzhou, China; [4] School of Information Engineering, Gansu University of Chinese Medicine, Gansu, Lanzhou, China; [5] School of Public Health, Gansu University of Traditional Chinese Medicine, Gansu, Lanzhou, China

第一机构：甘肃中医药大学

会议论文集：Proceedings of 2024 5th International Symposium on Artificial Intelligence for Medicine Science, ISAIMS 2024

会议日期：August 13, 2024 - August 17, 2024

会议地点：Amsterdam, Netherlands

语种：英文

外文关键词：Convolutional neural networks

年份：2025

摘要：To improve the accuracy of benign and malignant classification of lung nodules in CT images, a method based on feature fusion and improved random forest is proposed to classify benign and malignant lung nodules. First, the high-order features of lung nodules extracted by convolutional noise reduction self-encoder and convolutional neural network are fused; second, the random forest model is optimized by two-step chi-square test. Experiments on benign and malignant classification were conducted using 2000 lung nodule samples in the LIDC-IDRI dataset. The experimental results show that the proposed method has high performance in classifying benign and malignant lung nodules, with accuracy, sensitivity, and specificity of 0.9566, 0.9524 and 0.9626, respectively. Compared with the original random forest, the number of decision trees was reduced by 80% and the accuracy was improved by 0.0144. ? 2024 Copyright held by the owner/author(s).