| Objective:1.Evaluate the effectiveness of MaZda image texture analysis technology in the differential diagnosis of solitary pulmonary nodules.2.Analyze the clinical application value of MaZda image texture analysis technology in the diagnosis of solitary pulmonary nodules.Methods: 1.MRI images of 40 patients with solitary pulmonary nodules were retrospectively reviewed by nonenhanced MRI.The pulmonary granuloma(PG)and peripheral small lung cancer(SPLC)were classified into two groups according to the pathological results of surgery or biopsy.2.Using MaZda 4.6 software to extract the 6 image texture features of the three series of T1 WI,T2WI and DWI images by manually delineating the region of interest(ROI),including a first-order statistical parameters of histogram of the image gradient,such as: mean,variance,Second-order texture feature sunch as Contrast,Correlation,Entropy,and second-order texture generated by gray-level cooccurrence matrix(GLCM)methods.3.The statistical analysis of the texture features extracted from the MRI images in the lesion area was performed as follows.(1)The independent sample t-test in SPSS 24 software was used to compare and analyze the differences in image texture characteristics between the two groups.The results were average ± standard deviation((?)±s),t value,P value listed,P <0.05 was statistically significant.(2)A receiver operating characteristic curve(ROC)was established for the statistically significant difference in texture characteristics and their different combinations,and the area under the curve(AUC)was calculated.With 0.5<AUC<0.7 as the diagnostic efficiency,0.7 ~0.9 is the better diagnostic efficiency,and AUC>0.9 is the significant diagnostic efficiency.The threshold is determined according to the Youden index(Yorden index = sensitivity-(1-specificity)),and the specificity and sensitivity of differential diagnosis of benign granulomatous and peripheral small lung cancer are evaluated by evaluating texture parameters and their different combinations.The difference in AUC of each texture parameter and its different combinations was analyzed by Medcalc software,and P<0.05 was statistically significant.(3)The statistically significant difference between statistical features and their different combinations and manual interpretation was compared.Results:(1)Comparison of 5 texture parameters calculated by two statistical methods of T1 WI sequence:(1)The parameters of the texture features in the first-order histogram of the image: mean(T1grM): 3.58±0.74 for the isolated PG group and 3.06±0.71 for the SPLC group,the difference was statistically significant(t=2.278,P=0.028<0.05).Variance(T1grV): 6.68±2.59 in the isolated PG group and 4.59±1.32 in the SPLC group,the difference was statistically significant(t=3.174,P=0.004<0.05).(2)The parameters of the texture features in the second-order histogram GLCM of the image: contrast(T1Con): 0.44±0.12 in isolated PG group and 0.34±0.08 in SPLC group,the difference was statistically significant(t=3.164,P=0.003<0.05);correlation(T1Cor): isolated PG group 0.96±0.01,SPLC The group was 0.97±0.01,the difference was statistically significant(t=-3.24,P=0.002<0.05);entropy(T1Ent): 1.35±0.06 in the isolated PG group and 1.28±0.12 in the SPLC group,the difference was statistically significant(t= 2.333,P=0.025<0.05).(2)Comparison of 5 texture parameters calculated by two statistical methods of T2 WI sequence:(1)The parameter of the texture feature in the first-order histogram,mean(T2grM),Variance(T2grV),the difference was not statistically significant.(2)The parameters of the texture features in the second-order histogram GLCM of the image: contrast(T2Con),correlation(T2Cor),entropy(T2Ent),the difference was not statistically significant.(3)Comparison of 5 texture parameters calculated by two statistical methods of DWI sequence:(1)The parameter of the texture feature in the first histogram,the mean(DWIgrM): 5.01±1.01 for the isolated PG group and 4.22±0.84 for the SPLC group.The difference was statistically significant(t=2.726,P=0.01<0.05).DWIgrV: 7.56±4.92 in the isolated PG group and 5.42±2.31 in the SPLC group.The difference was not statistically significant(t=1.946,P=0.059).(2)The parameters of the texture features in the second-order histogram GLCM of the image: contrast(DWICon),correlation(DWICor),entropy(DWIEnt),the difference was not statistically significant.(4)Comparison of the AUC of each statistically significant texture parameter ROC and its different combinations:(1)Differential diagnostic performance of various texture features and their combinations: T1 grM,T1grV,T1 Con,T1Cor,T1 Ent in T1 WI sequence images and DWIgrM in DWI sequence images is the better diagnostic efficiency in differential diagnosis of solitary PG and SPLC(AUC > 0.7),T1 WI Statistically significant combination of gradient texture features(T1grMV),statistically significant combination of T1 WI GLCM texture features(T1CCE),statistically significant gradient and GLCM texture features of T1 WI texture features(T1MVCCE),and The combination of statistical features with T1 WI and DWI(T1-DWI)is the better diagnostic efficiency in the differential diagnosis of solitary PG and SPLC(AUC > 0.75).(2)The results of AUC difference between texture features and their combinations: There were no significant differences in AUC between the six texture features(P>0.05);the AUC difference between T1 grMV and 6 texture features and the other 3 texture combinations was not statistically significant.The significance(P>0.05);the difference of AUC between T1 CCE and 6 texture features and the other 3 texture combinations was not statistically significant(P>0.05);the difference of AUC between T1 MVCCE and T1 grM was statistically significant(z=2.009,P =0.0445,P<0.05),there was no significant difference in AUC between the other five texture features and the other three texture combinations(P>0.05);the difference in AUC between T1-DWI and T1 grM,T1Con,T1 Cor and T1 Ent was statistical significance,z = 1.983,2.061,1.998,2.191,P = 0.0473,0.0393,0.0458,0.0285,P < 0.05;T1-DWI and the other two texture features(T1grV,DWIgrM)and three other texture combinations,there was no significant difference in AUC between the groups(P>0.05).(5)Comparison with 6 texture features and 4 combinations of AUC:(1)There was no significant difference in AUC between 6 texture features(P>0.05).(2)There was no significant difference in AUC between texture feature combinations(P>0.05).Conclusions:(1)The texture features of T1 WI and DWI images in MRI conventional image texture analysis have certain value for the differential diagnosis of solitary PG and SPLC.The texture features contained in T2 WI images have certain difficulties in the differential diagnosis of solitary PG and SPLC.The texture features are yet to be further explored.(2)The texture features of T1 WI and DWI images in the MRI scan scan sequence are combined with the identification image,in the case of isolated PG and SPLC cases,the diagnostic performance of the combination of multiple texture features is better than the single texture feature,and the diagnostic performance of the multi-dimensional texture feature combination is better than the single-dimensional(first-order or second-order)texture features,and the multi-sequence texture combination requires better diagnostic performance than the single-sequence texture feature.(3)MRI conventional image texture analysis has certain scientific and clinical application potential in the interpretation of isolated PG and SPLC cases. |
PDF Full Text Request