The Diagnostic Performance Of Reduced Field-of-View Multi-b Diffusion-Weighted Imaging In Thyroid And Parathyroid Lesions

Objective①To preliminarily evaluate the diagnostic performance of quantitative diffusion parameters,namely IVIM derived pure diffusion coefficient(D),perfusion fraction(f)and pseudo-diffusion coefficient(D*),DKI derived mean kurtosis(MK)and mean diffusivity(MD),and DWI derived apparent diffusion coefficient(ADC),in terms of thyroid and parathyroid lesions using rFoV multi-b ZOOMit DWI.② To enrich the quantitative parameter database of normal thyroid parenchyma as well as thyroid and parathyroid lesions,and to provide high-resolution images which meet the demand of auto-recognition,auto-localization and auto-segmentation in order to lay the foundation for the application of artificial intelligence(such as deep learning,machine learning,etc.)in disease diagnosis.Method From March,2019 to October,2019,a total of 6 patients(2 males and 4 females;mean age,49.7±18.8 years;range,24-72 years)with thyroid and/or parathyroid lesions diagnosed by ultrasound,PET/CT or clinical blood test in Shandong Provincial Hospital Affiliated to Shandong University were prospectively enrolled in this study,and all of them had the will to undergo the MRI examinations.Meanwhile,6 healthy volunteers(3 males and 3 females;mean age,27.2±6.8 years;range,24-41 years)were enrolled as well.The MRI examinations were performed for all subjects at a 3T MRI scanner(MAGNETOM Prisma,Siemens Healthcare,Erlangen,Germany)with a 64-channel head-neck coil.Scanning parameters of the axial rFoV multi-b ZOOMit DWI sequence used for fitting IVIM and DKI models were as follows:FoV=120 mm×61.6 mm;ten b-values were used at 0,20,40,100,150,200,400,800,1000,and 2000 s/mm2,respectively.Scanning parameters of the axial rFoV ZOOMit DWI used for fitting conventional 2 b-values DWI model were as follows:FoV=120 mm×61.6 mm,two b-values(b1=0 s/mm2,1 average;b2=500 s/mm2,20 averages).For all 12 subjects,multi-b DWI data were processed and analyzed in Siemens Healthcare Syngo via Workstation.All diffusion parametric maps were calculated using a prototype post-processing software called Body Diffusion Toolbox(Version 1.3.0,Siemens Healthcare,Erlangen,Germany).The region of interest(ROI)was manually placed on each thyroid or parathyroid lesion(lesion group)and normal thyroid parenchyma(normal parenchyma group)at several different image segments.The value of D,f,D*,MK,MD and ADC calculated from each ROI in lesion group and normal parenchyma group was measured 3 times for average.Statistical analyses were performed using IBM SPSS Software(Version 26;Armonk,NY,USA).Measurement data were shown in the form of mean± SD(χ±s).The normal distribution and the homogeneity of variance of these 6 quantitative parameters were determined by the Kolmogorov-Smirnov test and F-test,respectively.According to test results,the Wilcoxon rank-sum test was used for comparing the D,f,D*,MK and MD values of lesions with normal parenchyma,and the two independent samples t-test(Student’s t-test)was used for comparing the ADC values between lesions and normal parenchyma.The receiver operating characteristic curve(ROC)analysis was used to determine the sensitivity,specificity and threshold value of D*,MK and ADC to differentiate between lesions and normal parenchyma.The areas under ROC curve(AUC)of D*,MK and ADC were calculated respectively and compared by Z-test.One way analysis of variance(ANOVA)was used to analyze the variance of D,f,D*,MK,MD and ADC values among subgroups of lesions and normal parenchyma.Statistical significance was defined by P<0.05.Results ①19 ROIs from 6 patients’ lesions were enrolled in lesion group,including 6 ROIs of anaplastic thyroid cancers(ATC),6 ROIs of parathyroid adenomas,5 ROIs of benign thyroid nodules and 2 ROIs of Hashimoto’s thyroiditis.53 ROIs of normal thyroid parenchyma form 6 volunteers and some patients who had localized lesions were enrolled in normal parenchyma group.②Significant differences were demonstrated in D*,MK and ADC values between lesion group and normal parenchyma group(Z=-2.166,P=0.030 for D*,Z=-2.166,P=0.030 for MK and t=-2.371,P=0.021 for ADC).D*,MK and ADC values of lesions were significantly lower than those of normal parenchyma.While there were no statistical differences of D,f,and MD values between lesions and normal thyroid tissues(Z=-0.134,P=0.893;Z=-0.722,P=0.470;Z=-0.287,P=0.774,respectively).③The AUC and its 95%CI of IVIM-D*value for distinguishing lesion from normal parenchyma was 0.668(0.497-0.839)(P=0.030),the sensitivity was 63.16%,the specificity was 86.70%and the diagnostic threshold of the D*value was 8.195×10-3mm2/s.The AUC with 95%CI of DKI-MK value was 0.668(0.497-0.839)(P=0.030),the sensitivity and specificity were 57.89%and 83.02%,respectively.And the diagnostic threshold of the MK value for distinguishing lesion from normal parenchyma was 1.047.The AUC with 95%CI of DWI-ADC value was 0.681(0.526-0.836)(P=0.028),its sensitivity and specificity were 76.47%and 59.57%,respectively.And the diagnostic threshold of the ADC value was 1.441×10-3mm2/s.There were no significant differences when making pairwise comparisons among the AUCs of D*,MK.④Malignant thyroid lesion group had significantly lower IVIM-f and DKI-MD values than benign thyroid lesion group(P=0.017 and P=0.000,respectively).Malignant thyroid lesion group had significantly lower DKI-MD values than parathyroid adenoma group(P=0.011).The IVIM-f,IVIM-D*,DKI-MK and DKI-MD values were significantly lower in malignant thyroid lesion group than those in normal parenchyma group(P=0.042,P=0.000,P=0.000 and P=0.002,respectively).The DWI-ADC values were significantly lower in parathyroid adenoma group than that in normal parenchyma group(P=0.029).There were no significantly statistical differences in D,f,D*,MK,MD and ADC values between benign thyroid lesion group and parathyroid adenoma group as well as benign thyroid lesion group and normal parenchyma group.Conclusion The rFoV ZOOMit DWI sequence has the capability to improve image quality in thyroid and parathyroid region.Parameter maps of IVIM,DKI and DWI models based on ZOOMit DWI are of relatively high quality with little deformation and artifacts,and thus enhance the accuracy of the quantitative measurement.In particular,the quantitative D*,MK and ADC values have better diagnostic performance than D,f and MD values for distinguishing abnormal thyroid and parathyroid lesions from normal thyroid parenchyma.However,the diagnostic efficiency of D*value is roughly equivalent to that of MK and ADC values.Moreover,parameters derived from IVIM,DKI and conventional DWI model are able to provide more biological information in differential diagnosis of malignant thyroid lesions,benign thyroid lesions,parathyroid lesions and normal thyroid tissues,which might be beneficial to the application of artificial intelligence aided diagnosis in this region.