Differential diagnosis efficiency between the benign and malignant solitary pulmonary nodules of the computer-aided diagnosis system with improved FCM algorithm

ZHANG Zewen; HUANG Chuanjun; ZHANG Xuan; ZHANG Caiqing

doi:10.20258/j.cnki.1006-9011.2025.06.014

Abstract：Objective To improve the fuzzy C-means(FCM)algorithm in the computer-aided diagnosis(CAD)system and to evaluate the auxiliary diagnosis efficiency of the CAD system with the improved FCM algorithm for benign and malignant soli-tary pulmonary nodules(SPN).Methods A total of 209 patients with SPN were enrolled and randomly divided into an experi-mental group(148 cases)and a validation group(61 cases).There were 74 benign SPNs and 74 malignant SPNs in the experi-mental group,and 30 benign SPNs and 31 malignant SPNs in the validation group.Chest CT images of 209 SPN patients were collected to extract lung parenchyma.The improved FCM algorithm was obtained by constructing a new two-dimensional vector and modifying the traditional objective function to segment SPN.The texture features of SPN in the experimental group were ex-tracted using the grey co-occurrence matrix in Matlab software.The four texture features,i.e.,correlation,entropy,homogeneity and energy,extracted from the experimental group were input into the support vector machine(SVM)model,and the SPN be-nign and malignant classification model was obtained after training.The construction of the CAD system with the improved FCM algorithm was completed.In the validation group,the sensitivity,specificity,accuracy,positive predictive value and negative predictive value of the CAD system with the improved FCM algorithm and the two radiologists using or not using the CAD system to diagnose the benign and malignant nature of SPN were compared to evaluate the effectiveness of the CAD system in indepen-dent diagnosis and combined diagnosis.Results The CAD system with the improved FCM algorithm was successfully con-structed.Based on the chest CT images of SPN patients,the sensitivity,specificity,accuracy,positive predictive value and negative predictive value of the CAD system with the improved FCM algorithm for diagnosing malignant SPN were 71.0%,70.0%,70.0%,71.0%and 70.0%,respectively.Compared with the diagnosis results of the radiologists alone,the sensitivity,specificity,accuracy,positive predictive value and negative predictive value of the two radiologists for diagnosing malignant SPN were improved after using the CAD system with the improved FCM algorithm.Conclusion Based on the chest CT images of SPN patients,the CAD system with the improved FCM algorithm can automatically distinguish the benign and malignant nature of SPN,and can improve the diagnostic efficiency of doctors in distinguishing the benign and malignant nature of SPN.