Study On The Value Of AI "pecking Doctors" In Differentiating Benign And Malignant Pulmonary Nodules

Objective: The analysis results of benign and malignant pulmonary nodules by AI "pecking doctors" were compared with the diagnosis results of intermediate radiologists using philips pulmonary nodules CAD software and manual analysis,to evaluate the diagnostic efficacy of the two in differentiating benign and malignant pulmonary nodules.Meanwhile,CT imaging characteristics of malignant pulmonary nodules were summarized and analyzed.Methods : A total of 210 pulmonary nodules with pathological results in 205 cases were collected: 1.The relationship was analyzed between the benign and malignant pulmonary nodules at age,sex,site distribution.2.According to the pathological types of lung nodules(benign lesion before,infiltration,infiltrating adenocarcinoma),nodule types(solid nodules,pure grinding glass nodules,mixed grinding glass nodules)and size(d≦1cm,1cm﹤d≦2cm,2cm﹤d≦3cm)into groups,analysed the accuracy of two ways in the diagnosis of pulmonary nodules,sensitivity,specificity,positive predictive value and negative predictive value,evaluated the value of pecked doctors and radiologists analysis the benign and malignant lung nodules.3.210 cases of pulmonary nodules by pathological type can be divided into benign lesion before,infiltrating group(not adenomatous hyperplasia and carcinoma in situ),infiltrating adenocarcinoma group three groups,analysed and recorded their gender,age,nodules imaging characteristics: parts,nodular type,size(length to diameter),volume,density,calcification,tumor-lung boundary,lobulation,spiculation,pleural depression,bronchography sign,vascular abnormality,cavitation performance,analysed the differences between the three groups.Results: 1.There was no significant difference in gender composition between benign and malignant groups(P>0.05);The mean age of the malignant group was higher than that of the benign group,and the difference was statistically significant(P<0.05).There was no significant difference in the distribution of pulmonary nodules between benign and malignant groups(P>0.05).Among 210 pulmonary nodules,57 were benign nodules and 153 were malignant nodules(21 were preinvasive lesions and 132 were invasive adenocarcinoma).A total of 54 were benign nodules and 141 were malignant nodules detected by AI(16 were preinvasive lesions and 125 were invasive adenocarcinoma),with a detection rate of 92.8%.The overall accuracy of AI in diagnosing pulmonary nodules was 68.2%(benign nodules 66.7%,preinvasive lesions 37.5%,and invasive adenocarcinoma 72.8%).The accuracy of AI diagnosis of solid nodules,pure ground glass nodules,mixed ground glass nodules and d≦1 cm,1cm ﹤ d ≦ 2cm,2cm ﹤ d ≦ 3cm nodules was 74.6%,45.8%,64.9%,55.6%,62.9%,80.3%,all lower than that of manual diagnosis.The sensitivity,specificity,positive predictive value and negative predictive value of AI in the diagnosis of pulmonary nodules were 68.7%,66.7%,84.3% and 45.0%,all lower than the manual diagnosis,especially the negative predictive value of nodules.However,the specificity of AI diagnosis of d 1cm nodules(95.0%)was higher than that of manual diagnosis(85%).The value comparison of AI in the diagnosis of other types of nodules was lower than that of manual diagnosis,especially the sensitivity of pure ground glass nodules and subcm and the specificity of d>1cm nodules,but the positive predictive value of mixed ground glass nodules and d>1cm nodules was close.3.There was no statistically significant difference in tumor-lung boundary between the three groups(P>0.05).There were statistically significant differences in the type,size,volume,density,calcification,lobulation,speculation,pleural depression,bronchography sign,vascular abnormalities,and vacuolar signs between the three groups(all P<0.05).Conclusion: 1.The older the patient,the higher the probability of nodule malignancy.2.The current version of AI "pecking doctors" is generally less valuable for the differential diagnosis of benign and malignant pulmonary nodules than manual diagnosis,but it is more specific for the diagnosis of pulmonary nodules with d<1cm,which is beneficial for radiologists to judge the benign and malignant pulmonary nodules with subcm,and has a good auxiliary role in the diagnosis of other types of nodules and has certain clinical value.3.With the increase of pulmonary nodules,the probability of malignancy increases.Meanwhile,with the increase of pulmonary nodules and its density,the infiltration of pulmonary nodules may increase.The pulmonary nodules’ lobulation,spiculation,pleural depression,bronchiolography sign,vascular abnormality and vacuolation all indicate the possibility of malignant nodules.The presence of lobulation,spiculation and pleural depression indicates the invasiveness of nodules,and the presence of vascular abnormality and vacuolation is helpful for the determination of malignant nodules.