An Apparent Diffusion Coefficient Histogram Method Versus A Traditional Two-dimentional Measurement Method For Identifying Non-puerperal Mastitis From Breast Cancer

BACKGROUND:There is overlap between non-puerperal mastitis and breast cancer in clinical manifestations and conventional imaging features so that they tend to be mixed up.What's more,regarding whether ADC values can identify the above two breast lesions or not is still controversial in clinical studies.OBJECTIVE:The aim of this study was to investigate the utility of a volumetric ADC histogram method at 3.0T for distinguishing NPM from BC,and to compare this method with a traditional two-dimensional measurement method,that is small solid sample method.MATERIAL AND METHODS:Twenty-seven patients with non-puerperal mastitis and fifty-three patients with breast cancer who met the inclusion and exclusioncriteria withoutreceivinganyoperation,radiotherapyor chemotherapy and other breast-related therapy underwent conventional plain,dynamic contrast enhancement and diffusion weighted imaging?b=0,800s/mm²?scans.Apparent diffusion coefficient maps were obtained through assorted post-processing working station ADW 4.6,two radiologists independently measured ADC values according to two distinct region of interest?ROI?protocols.The first protocol?namely whole-volume ADC histogram method?included the generation of ADC histograms for each lesion and a various of parameters were examined,including the mean,minimum,percentiles(5th,10th,25th,50th,75th,90th,and 95^th),maximum,mode,skewness,and kurtosis,by delineating the boundaries of breast lesions taking T2-weighted imaging and T1-weighted dynamic contrast enhancement imaging as references.In the second protocol?namely small solid sample method?,three freehand?TF?ROIs for local lesions were generated to obtain a mean ADC value(defined as ADC_TF).Three circular or oval ROIs were placed where DWI visually presented hyperintense signal and ADC map showed low signal intensity,the diameter of each ROI was 5 to 10mm,avoiding necrosis,cyst degeneration,hemorrhage,calcifications.All the ADC values were statistically analyzed,a two-tailed P-value<0.05 was considered to be statistically significant.In addition,receiver operating characteristic?ROC?analyses were conducted for the significant parameters to assess the ability to identify NPM from BC.Corresponding areas under the ROC curve?AUC?,the threshold value,Youden index,sensitivity,specificity,positive predictive value?PPV?,negative predictive value?NPV?and accuracy were computed.RESULTS:1.Regarding the parameters for the whole-volume ADC histogram method:?1?Seen from statistics results,mean,maximum,mode,and all percentile ADC values were found to be significantly lower in the BC cases than in the NPM cases?all P<0.05?except for minimum,skewness,and kurtosis ADC values?all P>0.05?.?2?The results of the receiver operating characteristic?ROC?analysis of significant ADC histogram parameters showed that the 95^thh percentile ADC value yielded the highest AUC values?0.821;95%confidence interval[CI]:0.723,0.895?,at a cutoff value of 1.68×10^-33 mm²/s,the sensitivity,specificity,PPV,NPV,and accuracy values were:81.48%,78.12%,78.83%,80.84%,and 79.80%,respectively.Although there were no significant differences between the AUC values at the 90^thh and 95^thh percentiles?P=0.120?,the 90^thh percentile cutoff ADC value yielded the second highest AUC values?0.805;95%confidence interval[CI]:0.706,0.883?,at a cutoff value of 1.49×10^-33 mm²/s,the sensitivity,specificity,PPV,NPV,and accuracy values were:68.52%,84.37%,81.43%,72.83%,and 76.45%,respectively.?3?The leave-one-out cross validation?LOOCV?analysis indicated that the 90^thh percentile cutoff ADC value provided greater diagnostic accuracy than the 95^thh percentile cutoff ADC value?75.6%vs.50.0%,respectively?,and a combination of both percentile ADC values increased the classification accuracy to 97.7%.2.Regarding ADC_TFF for a traditional two-dimensional measurement method:ADC_TFF did not significant differ between the NPM and BC cases?P>0.05?.CONCLUSIONS:Whole-volumeADChistogramintroducesa multi-parametric quantitative diagnostic method.The generation of volumetric ADC histograms appears to be a superior method to the traditional two-dimensional method that was examined,and it also appears to represent a promising image analysis method in distinguishing the property of breast lesions.OBJECTIVE: To analyze the measurement repeatability of breast-related quantitative ADC histogram parameters.MATERIAL AND METHODS: Eighty-six breast lesions?including 32non-puerperal mastitis and 54 breast cancer?meeting the inclusion and exclusion criteria as stated in Part One were sequentially enrolled to analyze the measurement repeatability of breast-related quantitative ADC histogram parameters.Pretreatment ADC maps were evaluated independently by two radiologists who have more than five years of diagnostic experiences over magnetic resonance imaging in breast disease,and these two radiologists were unknown to the enrolled patients' clinical data,pathology result and treatment program.Taking T2-weighted imaging and dynamic contrast enhancement T1-weighted imaging with fat suppression as references,the boundary of breast lesion on each scan level was outlined,be careful to exclude normal tissue areas and obvious artifacts in the designation of ROIs.ADC histogram of breast lesion on each scan level was obtained by an in-house software?Image J,NIH,Bethesda,MD,USA?,frequency distribution concerning ADC values was presented in the Excel table.Export and gather the data in Excel table.And then a volumetric ADC histogram equivalent to the cumulative ADC values of each voxel of the whole lesion was generated by using SPSS 22.0?Chicago,IL,USA?statistical software.Therefore,a series of parameters,including the mean,minimum,percentiles?5th,10 th,25th,50 th,75th,90 th,and 95th?,maximum,mode,skewness,and kurtosis were measured.Inter-reader consistency between the two radiologists was measured by using an intra-class correlation coefficient?ICC?and its 95% confidence interval?CI?.The ICC was a dynamic target that ranged from 0 to 1,with ICC values closer to 1 indicating better reproducibility.ICCs were evaluated as follows: minor reproducibility?0–0.20?,fair reproducibility?0.21–0.40?,moderate reproducibility?0.41–0.60?,substantial reproducibility?0.61–0.80?,and excellent reproducibility?0.81–1.00?.RESULTS: In general,the inter-reader consistencies between the two radiologists for the ADC histogram parameters were excellent for the mean,minimum,maximum,and all of the percentile ADC values?ICC range,0.81–0.95?.Among these,the 50 th percentile values exhibited the highest inter-reader consistency?ICC,0.95?.Meanwhile,the inter-reader consistencies for the ADC histogram parameters were substantial for the mode and skewness?ICC=0.80 and 0.70,respectively?.However,inter-reader consistency regarding kurtosis was just moderate?ICC = 0.57?.CONCLUSIONS: Measurements of the mean,minimum,maximum,and all of the percentile ADC values from breast-related ADC histogram were confirmed to be repeatable.However,the measurement of kurtosis may bring about bias.