The Preliminary Research Of Diffusion Weighted Imaging Based On Monoexponential And Biexponential Model In The Differential Diagnosis Between Benign And Malignant Lung Nodules/Masses

This topic is divided into two parts: Part I: the effect of region of interest(ROI)selection on the consistency of lung nodules/masses measurement in diffusion weighted imaging(DWI)based on monoexponential and biexponential model;Part II: clinical study in DWI based on monoexponential and biexponential model in the differential diagnosis between benign and malignant lung nodules/masses.Part I: The effect of ROI selection on the consistency of lung nodules/masses measurement in DWI based on monoexponential and biexponential modelObjective Comparison of each parameter in interobserver’s and intraobserver’s consistency and repeatability,obtained from DWI based on monoexponential and biexponential model,to choose stable,repeatable and simple ROI method to provide technical support for differential diagnosis of benign and malignant lung lesions in part Ⅱ.Methods 43 lung nodules/masses underwent conventional DWI(b=0,800s/mm2)and multi-b values DWI(b=0,50,100,150,200,250,300,500,800,1000s/mm2)by 3.0T DWI.Two observers blindly used fixed area of ROI(ROIf)to measure singal intensity(SI)of spinal cord and pulmonary nodule/mass,then measure the lesion-to-spinal cord ratio(LSR),used maximunm area of ROI(ROIm),round area of ROI(ROIr)and volume of ROI(ROIv)to measure ADCmin and ADCmean of lung nodule/mass,used ROIm and ROIv to measure f,D,D* of lung nodule/mass.The intra-class correlation coefficient(ICC)was determined for inter-observer and intra-observer analysis,while the Bland-Altman plot was determined to analysis the dots’ distribution.Besides,the well repeatable parameters having same nature obtained from different ROIs were compared,and P<0.05 was considered statistically significant.Result 1.LSR(ROIf)and ADCmean(ROIm/ROIr/ROIv)had a high consistency regardless of interobserver and intraobserver(ICC: 0.762~0.979,within subject coefficient of variation(WCV): 2.47%~9.79%,while ADCmin(ROIm/ROIr/ROIv)had a poor reproducibility both in interobserver and intraobserver(ICC: 0.492~0.724,WCV: 18.44%~28.88%).2.f,D,and D*,produced by both ROIm and ROIv,showed a high consistency in interobserver and intraobserver(ICC: 0.765~0.995,WCV: 2.47%~18.39%).3.In monoexponential model DWI,ADCmean(ROIm)was not statistically different from ADCmean(ROIr)and ADCmean(ROIv)(P>0.05),while ADCmean(ROIr)was statistically different from ADCmean(ROIv)(P<0.05).In biexponential model DWI,D and D* from two ROIs(ROIm,ROIv)had no significant difference(P>0.05),however,f(ROIm)was statistically different from f(ROIv)(P<0.05).Conclusion 1.The values of LSR(ROIf),ADCmean(ROIm/ROIr/ROIv),f(ROIm/ROIv),D(ROIm/ROIv)and D*(ROIm/ROIv)were stable,reliable and repeatable.2.The repeatability of ADCmin(ROIm/ROIr/ROIv)was very low.3.ADCmean obtained by ROIr and ROIv were statistically significant,which meaned ADCmean obtained from different ROIs couldn’t replace each other.4.f obtained by ROIm and ROIv were statistically significant,which meaned f obtained from different ROIs couldn’t replace each other.Part II: Clinical study in DWI based on monoexponential and biexponential model in the differential diagnosis between benign and malignant lung nodules/massesObjective Each parameter’s diagnostic efficacy was compared to provide the basis for clinician selection of appropriate,effective ROI method and its parameter and corresponding threshold in the differential diagnosis between benign and malignant lung nodules/masses.Methods 43 lesions showed as lung nodules/masses by CT were retrospectively analyzed.They were categorized into benign(14 lesions)group and malignant(29 lesions)group by histopathology,needle biopsy or clinical follow-up.All patients underwent conventional DWI(b=0,800s/mm2)to obtain LSR(ROIf)and ADCmean(ROIm/ROIr/ROIv),multi-b values DWI(b=0,50,100,150,200,250,300,500,800,1000s/mm2)to obtain f(ROIm/ROIv),D(ROIm/ROIv),and D*(ROIm/ROIv)by 3.0T DW-MRI.The above parameters were compared between the benign group and malignant group,respectively.P <0.05 was considered statistically significant.Besides,the receiver operating characteristic curve(ROC)was drawn to analyze the diagnostic efficacy of meaningful parameters.Result 1.LSR(ROIf),ADCmean(ROIm/ROIr),f(ROIm),D(ROIm)and f(ROIv)were statistically significant in the differential diagnosis between benign and malignant lung nodules/masses(P<0.05).However,ADCmean(ROIv),D*(ROIm),D(ROIv)and D*(ROIv)were not statistically significant.2.The area under curve(AUC)of LSR(ROIf),ADCmean(ROIm),ADCmean(ROIr),f(ROIm),D(ROIm)and f(ROIv)were 0.887,0.803,0.746,0.754,0.697 and 0.596,respectively.Their corresponding best threshold in differential diagnosis benign and malignant lung lesions were 1.655,1.402×10-3mm2/s,1.488×10-3mm2/s,18.365%,1.303×10-3mm2/s,27.340%,respectively.The sensitivity of above parameters in diagnosi s of pulmonary malignant lesions were 75.9%,75.9%,86.2%,58.6%,75.9%,75.9%,respectively,the specificity were 92.9%,78.6%,71.4%,85.7%,57.1%,64.3%,respectively.Conclusion 1.DWI based on monoexponential and biexponential model both could differentiate the benign and malignant lung nodules/masses,and the diagnostic efficacy of LSR was the best.2.ADCmean(ROIv),D*(ROIm),D(ROIv)and D*(ROIv)were meaningless in the differential diagnosis between benign and malignant lung nodules/masses.3.ROI could affect ADCmean and D in the differential diagnosis between benign and malignant lung nodules/masses,so clinical application of ROI should be especially noted.