The Application Of Mammography Detection System Based On Artificial Intelligence In Diagnosis Of Breast Tumors

Objective:This thesis is to investigate the application value of the mammography detection system based on artificial intelligence in the diagnosis of breast tumors.Methods:Research data:1.The data of this thesis included validation set and test set.The validation set was used to adjust parameters and optimize the system of MammoWorks^TMM detection,the test set was used to test the optimized system.2.A total of 1,413 cases were enrolled,of which 1,218 were validation set and 195were test set.3.448 cases with BI-RADS 5⁶ and 448 cases with BI-RADS 1 were included in the validation set,as well as 322 cases of nonpalpable breast tumors.4.130 cases with BI-RADS 5⁶ and 65 cases with BI-RADS 1 were included in the test set.Research methods:To analyse mammography images of all cases by using MammoWorks^TMM breast intelligent detection system.1.The investigation of the diagnostic value of MammoWorks^TM（versions 3.5.2.6）in the validation set1)The detection efficacy of MammoWorks^TM（version 3.5.2.6）in breast cancer,and the effect of clinicopathological characteristics on it.2)The detection efficacy of MammoWorks^TM（version 3.5.2.6）in nonpalpable breast tumors,and the effect of clinicopathological characteristics on it.2.The investigation of the diagnostic value of MammoWorks^TM（versions 3.5.4.43）in the validation set1)The detection efficacy of MammoWorks^TM（version 3.5.4.3）in breast cancer,and the effect of clinicopathological characteristics on it.2)The detection efficacy of MammoWorks^TM（version 3.5.4.43）in nonpalpable breast tumors,and the effect of clinicopathological characteristics on it.3.Thedifferenceofdetectionefficacybetweendifferentversionsof MammoWorks^TMM in the validation set1)The detection difference between different versions of MammoWorks^TMM in the validation set.2)The difference of false-positive markers in one mammography between different versions of MammoWorks^TMM in the validation set.4.The evaluation of the diagnostic efficacy of MammoWorks^TM（version 3.5.4.43）by using the test set data.Statistical methods:To statistical analyzed the sensitivity,the specificity,the accuracy,the positive predictive value,the negative predictive value,the positive likelihood ratio,the negative likelihood ratio and the false-positive markers in one mammography of MammoWorks ^TM（version 3.5.2.6 and version 3.5.4.43）by SPSS20.0.The rate comparison was performed using?² test,corrected?² test,and Fisher exact probability method.The false-positive markers in one mammography were compared by non-parametric test.The consistency of different subgroups was evaluated by Kappa test.Results:1.The application of MammoWorks^TM（version 3.5.2.6）in the validation set1)The accuracy of MammoWorks^TM（version 3.5.2.6）in breast cancer was 47.76%.The sensitivity was 74.33%and the specificity was 21.21%.The positive predictive value was 48.54%and the negative predictive value was 45.24%.The positive likelihood ratio was 0.94 and the negative likelihood ratio was 1.21.The false-positive markers in one mammography were 0.50（0.25⁰.75）.The sensitivity of the MammoWorks^TMM system showed statistically significant difference between craniocaudal（CC）view and mediolateral oblique（MLO）view of mammography,and the sensitivity was statistically significant different in patients with different tumor sizes,breast densities,and lesion types（P<0.05）.The false-positive markers in one mammography showed statistically significant difference between the diseased group and non-diseased group,and in patients with different breast densities（P<0.05）.2)The accuracy of MammoWorks^TM（version 3.5.2.6）in nonpalpable breast tumors was 78.57%,the sensitivity 78.57%,the positive predictive value（the number of markers）55.32%.The false-positive markers in one mammography were 0.50（0.25⁰.75）.The sensitivity was statistically significant different in patients with different breast densities,and different lesion types（P<0.05）.The false-positive markers in one mammography showed statistically significant difference in different breast density groups（P<0.05）.2.The application of MammoWorks^TM（version 3.5.4.43）in the validation set1)The accuracy of MammoWorks^TM（version 3.5.4.43）in breast cancer was 52.12%.The sensitivity was 94.64%and the specificity was 9.6%.The positive predictive value was 51.15%and the negative predictive value was 64.18%.The positive likelihood ratio was 1.05 and the negative likelihood ratio was 0.56.The false-positive markers in one mammography were 1.00（0.50¹.50）.There was no significant difference in different clinicopathological groups（P>0.05）.The false-positive markers in one mammography showed statistically significant difference in different breast density groups（P<0.05）.2)The accuracy of MammoWorks^TM（version 3.5.4.43）in nonpalpable breast tumors was 95.65%,the sensitivity 95.65%,the positive predictive value（the number of markers）44.27%.The false-positive markers in one mammography were 1.00（0.75¹.50）.There was no significant difference in different clinicopathological groups（P>0.05）.The false-positive markers in one mammography showed statistically significant difference in groups of benign,malignant cancer and precancerous lesions,as well as in different breast density groups（P<0.05）.3.The difference of detection efficiency between different versions of MammoWorks^TMM in the validation set1)The detection accuracy,the sensitivity,the positive predictive value,the negative predictive value,and the positive likelihood ratio of breast cancer were increased after the updating of the system,and the negative likelihood ratio and the specificity were decreased.The detection accuracy and the sensitivity of nonpalpable breast tumors were increased,the positive predictive value was decreased（P<0.05）.The detection efficiency in different clinicopathological groups was higher than before（P<0.05）.2)Compared with the previous version,the false-positive markers in one mammography increased in diseased group,non-diseased group,and nonpalpable breast tumor group（P<0.05）.4.The detection efficacy of MammoWorks^TM（version 3.5.4.43）was evaluated by the test set.The accuracy was 68.21%.The sensitivity was 86.15%and the specificity was32.31%.The positive predictive value was 71.79%and the negative predictive value was53.85%.The positive likelihood ratio was 1.27 and the negative likelihood ratio was 0.43.The false-positive markers in one mammography were 0.75（0.25¹.00）.There was no statistically significant difference on the detection efficiency and the false-positive marks in one mammography in different clinicopathological groups（P>0.05）.Conclusion:1.The detection sensitivity of MammoWorks^TMM system gets a certain degree of improvement after optimizing,and it is helpful in the auxiliary diagnosis of breast tumors.2.MammoWorks^TMM system has certain application in the early diagnosis of breast cancer after updating.3.Because of the low specificity and the more false-positive markers in one mammography,the MammoWorks^TMM system still needs to be optimized,and the detection efficiency still needs to be improved.