Diagnosis and analysis of pulmonary nodules using AI technology combined with multi-slice spiral CT

YE Hongda; WU Ligong; ZHOU Rong; MA Xiaodong; ZHENG Qingxiu; XIA Shengwei

doi:10.20258/j.cnki.1006-9011.2025.10.016

Abstract：Objective To analyze the diagnostic value of artificial intelligence(AI)technology combined with multi-slice spiral CT for pulmonary nodules.Methods A retrospective analysis was conducted on the clinical data of 98 patients with pul-monary nodules in the hospital,all of whom underwent multi-slice spiral CT examination.AI technology,physician manual re-view time,and lung nodule detection status were compared.Pathological examination results were used as the"gold standard"to compare the consistency between AI technology,physician manual diagnosis,and AI technology combined with physician manual diagnosis of lung nodule properties with pathological examination results.Results The average reading time per case of AI technology was shorter than that of physician manual(P<0.05).A total of 335 true nodules were detected in 98 patients,in-cluding 107 pulmonary nodules,with a diameter<5 mm,172 pulmonary nodules,with a diameter of 5～10 mm,and 56 pulmo-nary nodules,with a diameter>10 mm.The detection rate of<5 mm pulmonary nodules,5～10 mm pulmonary nodules,and the total detection rate of pulmonary nodules using AI technology were higher than those manually performed by physicians(P<0.05).Pathological examination results confirmed that out of 335 true nodules,130 were malignant lung nodules and 205 were benign lung nodules.According to the Kappa consistency test,the consistency between the results of AI technology and physician manual diagnosis of lung nodule properties and pathological examination results was generally average(Kappa was 0.611,0.665,all P<0.001);the consistency between the results of AI technology combined with physician manual diagnosis of lung nodule properties and pathological examination results was strong(Kappa=0.841,P<0.001).The accuracy of AI technology com-bined with physician manual diagnosis of lung nodule properties was higher than that of physician manual diagnosis and AI tech-nology,the sensitivity was higher than that of physician manual diagnosis,the specificity was higher than that of AI technology,the missed diagnostic rate was lower than that of physician manual diagnosis,and the misdiagnosis rate was lower than that of AI technology.The sensitivity and misdiagnosis rate of AI technology in diagnosing the nature of pulmonary nodules were higher than those of manual diagnosis by physicians,while,the specificity and missed diagnostic rate were lower than those of manual diag-nosis by physicians(P<0.05).Conclusion The combination of AI technology and multi-slice spiral CT can shorten the reading time and improve the detection rate of pulmonary nodules.However,AI cannot be directly used to determine the nature of pulmo-nary nodules and can be used as an auxiliary tool for manual detection to improve diagnostic accuracy.