Analysis of Accuracy and Consistency of AI-assisted CT Imaging in Diagnosing Pulmonary Nodules

Wu Yuping; Du Jingfang; Li Ya

doi:10.3969/j.issn.1004-5775.2025.17.021

Abstract：Objective:To analyze the accuracy and consistency of computed tomography(CT)imaging aided by artificial intelligence(AI)in the diagnosis of pulmonary nodules.Methods:Retrospective analysis was performed on the clinical data 50 patients with suspected pulmonary nodule in the hospital from January 2023 to November 2024.All patients underwent CT examination and CT+AI examination,and the relevant data were analyzed using surgical pathology as the gold standard.Results:Among 50 suspected pulmonary nodules,13 were positive and 37 were negative.Among them,3 cases were malignant,including 1 case of squamous cell carcinoma,1 case of carcinoid and 1 case of small cell carcinoma.Among the 10 benign cases,there were 3 nodules,2 nodules,2 adenomas,1 inflammatory myofibroblastoma and 2 sclerosing hemangioma.CT diagnosis results:11 cases were positive,6 cases were at moderate risk(5 cases at moderate risk,1 case at high risk),and the consistency with pathological diagnosis was general(Kappa=0.658).CT+AI diagnosis results:13 cases were positive,12 cases were at moderate risk(8 cases at moderate risk,4 cases at high risk),and the consistency with pathological diagnosis was good(Kappa=0.978).The accuracy and sensitivity of CT+AI were higher than those of CT(χ2=6.500,8.306;P＜0.05).There was no significant difference in specificity between two examinations(χ2=2.902,P＞0.05).In the nodule characteristics,the detection rates of nodules with diameter of 5 mm,6-7 mm,8-10 mm,solid nodules,ground glass nodules and middle pleura were higher than those of CT+AI(χ2=4.667,7.619,6.923,4.886,4.267,5.283;P＜0.05).There was no significant difference in the nodular diameter of 11～15 mm,partial firmness,detection rate of surrounding and other locations(χ2=1.400,1.243,1.243;P＞0.05).The average reading time of CT+AI was shorter and the image quality score was higher(t=19.418,5.963;P＜0.05).Conclusion:AI-assisted CT imaging diagnosis shows good accuracy and consistency in the detection and classification of pulmonary nodules,especially in the recognition ability of small nodules.