| Part 1Multi-detector computed tomography features implications of pure ground-glass noduleObjectivesThis study aimed to evaluate the CT features of pGGNs to identify factors predictive of pathological invasion by correlating the CT imaging features of persistent p GGO with pathological changes.Materials and MethodsWe retrospectively evaluated 104 pulmonary pGGNs from 102 patients selected between November 2013 and October 2015.All lesions have pathology results.pGGNs are divided to 3 groups(preinvasive group,MIA group and I-ADC group)according to differente invasive extent.Patients’ clinnical information and CT appearances were analyzed including age,gender,size,density,shape,margin,tumour-lung interface,internal and surrounding malignant signs.Computed tomographic scans in patients with pGGNs associated with airspaces were reviewed,pGGNs were divided to two groups,preinvasive and MIA group,I-ADC group;pGGNs’ size,multiplicity,distribution,location were analysed.All patients underwent multi-detector computed tomography(MDCT)examinations,then the 1 mm or 0.625 mm thin slices were obtained by workstation for statistical analysis.Results1.Statistics of Clinical Data104 pGGNs on CT in 102 patients with pGGNs were selected from november 2013 and October 2015.All lesions have pathology results.Size,density,tumour-lung interface,shape,margin,,internal and surrounding malignant signs of pGGNs wre analyzed.31 males and 40 females with an average age of 53.43(±10.43)were collected,Age ranged from 27~78;pGGNs divided to 3 groups,preinvasive group44(AAH 19,AIS 25),MIA group(MIA 31)and I-ADC group(I-ADC 29).24 pGGNs associated with airspaces were abtained,patients include 9 males and 15 females,age ranged from 29~69,pGGNs were divided to two groups,preinvasive and MIAgroup(AAH 1,AIS 4,MIA 8),I-ADC group(I-ADC 11).2.The CT Features and the Clinical information statistics of pGGNsNo significant differences were found in age,gender,density and shape air-bronchogram between three groups(P=0.313,P=0.566,P=0.108,P=0.194);It is of significant differences in size,lobulation,cystic airspaces,tumour-lung interface,pleural indentation and vascular change between 3 groups(P=0.000,P=0.040,P=0.005,P=0.028,P=0.023,P=0.003).Tumor size was significantly larger in the I-ADC groups than in the preinvasive group and MIA group(P=0.000;P=0.008).ROC curve analyses of tumor size shows that,the areas under the curve(AUC)was 0.75,the sensitivity and specificity of tumor size were 86.2% and 46.7%,respectively,the cutoff point is 1.05 cm.3.The CT Features of airspaces of pGGNsNo significant differences were found in multiplicity,distributiong of airspaces bigger than 5mm(P=0.816,P=0.608)was found between two groups;It is of significant differences in size and location(P=0.018,P=0.047)between two groups;ROC curve analyses shows that,the areas under the curve(AUC)was 0.78,the sensitivity and specificity of tumor size were 100% and 61.5%,respectively,the cutoff point is 2.7mm.Conclusions1.Tumor size,lobulation,tumour-lung interface,cystic airspaces,pleural indentation and vascular change can help predict invasive extent of early stage lung adenocarcinoma with p GGO.2.Airspace size,location can help differientiate I-ADC from MIA and preinvasive with pGGNs..Part 2 Computerized Texture Analysis of Pure Ground-Glass Nodules:Differentiation of Invasive Pulmonary Adenocarcinomas From Preinvasive Lesions or Minimally Invasive AdenocarcinomasObjectivesWe aimed to analysis the ability of computed tomography texture analysis to differentiate pure ground-glass nodules(PGGNs)confirmed to be invasive pulmonary adenocarcinomas(IPAs)from preinvasive lesions and minimally invasive adenocarcinomas(MIAs).Materials and MethodsThis retrospective study included 102 patients(41 men and 61 women)with 104 PGGNs on unenhanced computed tomography pGGNs selected between November 2013 and October 2015.All pGGNs have pathology results.PGGNs were categorized into 2groups,IPAs group and preinvasive group;Every pGGN was manually semi-automatic segmented and their computerized texture features were extracted by a CAD(Computer assistant diagnose)software program quantitatively.Mann Whitney U test,binary logistic regression and ROC curve analyses were performed to find differentiating factors of invasive pulmonary adenocarcinomas from preinvasive lesionsResults1.In 104 pGGNs from 102 patients,94 pGGNs from 94 patients were successfully segmented.PGGNs were categorized into 2 groups,IPAs group(n=55,MIA 30,I-ADC 25)and preinvasive group(n=39,AAH 16,AIS 23).2.No significant differences have been found in kewness,Kurtosis(First-order texture features),Entropy,Contrast,Correlate(Second-order texture features,GLCM matrix)between two groups.It is of significant differences in Mean CT number,Standard deviation of First-order texture features,Volume,Effective diameter,Surface area,Mass of Volumetric parameters,Energy,IDM(inverse different moment)of Second-order texture features(GLCM matrix)between two groups.3.As results of binary logistic regression analysis,larger nodule mass [adjusted odds ratio(OR),1.005] and higher IDM(adjusted OR,1.044)are differential factors of invasive pulmonary adenocarcinomas.A Binary logistic regression model using these two features showed excellent and significantly higher differentiating performance comparedto mass or IDM alone.ROC curve analyses shows that,as regard to combination,the areas under the curve(AUC)was 0.81,the sensitivity and specificity of tumor size were 100%and 87.3%,respectively,the sensitivity and specificity as regard to mass were 92.7% and64.1%,respectively,the areas under the curve(AUC)was 0.76;the sensitivity and specificity as regard to IDM were 83.6% and 51.3%,respectively,the areas under the curve(AUC)was 0.74.Conclusions1.Computed tomography texture features such as higher mass and IDM were significant differentiating factors of IPAs presenting as PGGNs.2.Preinvasive lesions can be accurately differentiated from IPAs by using computerized texture analysis. |
PDF Full Text Request