The Study Of Full-field Digital Mammography,ultrasonography And MRI In Diagnosing Breast Cancer

Part 1 ROC curve Study of full-field digital mammography, ultrasonography and MRI in diagnosing breast cancerObjective: To assess full-field digital mammography, ultrasonogrphy and MRI diagnositic value of breast cancer by ROC curve, and to analyze the causes of misdiagnosis and missed diagnosis for three modalities.Materials and Methods: 57 patients with 60 malignant lesions were enrolled during December 2007 to December 2009 in our hospital, which were all performed with mammography, ultrasonography and MRI. All the subjects were confirmed by pathology. In addition, 23patients with 24 lesions were also enrolled as the control group, all of them were performed with mammography, ultrasonography and MRI. The patients were examined with conventional cranicaudal, and mediolateral oblique using Siemens Mammomat Novation DR. If necessary, medial-lateral was checked. The ultrasonography imaging was performed with GE LOGIQ 9 connected to a wide frequency linear array probe, from 7.5~12MHZ. Color Doppler Flow imaging was used to observe the blood flow of the lesions. All the patients were performed with Siemens Magnetom Avanto 1.5T magnetic resonance scanner. Breast surface coil was used. Firstly, unenhanced scanning and diffusion weighted imaging of chest were done. Seconedly, unenhanced scanning, diffusion weighted imaging and dynamic contrast enhanced imaging of the breast were done. According to the U.S. Institute of Radiology breast imaging reporting and data system (BI-RADS, 2003) classi- fication and the grade of suspicion of malignancy by imaging manifestations, every lesion was classified by mammography, ultrasonography and MRI as follows: categoryⅠ: certainly benign lesions; categoryⅡ: probably benign lesions; categoryⅢ: qualitative should not; categoryⅣ: probably malignant; categoryⅤ: definitely malignant. The imaging classifications were correlated with pathology diagnosis. The statistic analysis of data was performed with SPSS 13.0, p values smaller than 0.05 were considered significant in all cases.Results: 1. The detection rate of the malignant group's calcification for mammography was higher than ultrasonography, the discrepancy had statistic significance (P<0.005); mammographic detection rate for the malignant calcified focus was 46.7%, ultrasonographic was 23.3%. MRI did not detect calcification in all cases. 2. Of the 60 breast malignant lesions, the mammographic classification was: category I: 4 lesions, category II: 2, category III, IV andⅤ: 1, 16, 37, respectively. The sonographic and MRI classification were: 1,1,7,21,30; 0,1,1,13,45, respectively。This study shows that AX线= 0.878 (P=0.000) , A超声= 0.865 (P=0.000) and A MRI = 0.925 > 0.9(P=0.000). MRI for the diagnosis of breast lesions had a significant statistical significance, diagnostic performance and higher, that mammography and ultrasonography for the diagnosis of breast lesions have a significant statistical significance, diagnostic performance and moderate. The diagnostic value of MRI was higher ultrasonography, and they had a significant statistical significance(P<0.05).But the curves of mammography and others modalities intersected and didn't compare simply.Conclusion: 1. For the malignant calcified focus detection rate, mammography was highest, ultrasonography was secondary, MRI was worest. 2. MRI for the diagnosis of breast lesions had a higher diagnostic performance, mammography and ultrasonography for the diagnosis of breast lesions had a moderate diagnostic performance. The diagnostic value MRI was higher ultrasonography. 3. Three modalities all had difficulties for some breast lesions, when there were difficult for diagnosing malignant and benign lesions, we need to combine other modality in order to elevate accuracy. Part2 The value of full-field digital mammography, ultrasonography and MRI in the assessment of breast cancer sizeObjective: To evaluate the value of mammography, ultrasonography and MRI to predict the histological invasive tumor size pre-operatively; and assess the possible causes of over- and underestimation of lesion size on mammography, ultrasonography and MRI.Materials and Methods: 46 patients with 46 malignant lesions were enrolled from the first part patients which were all performed with mammography, ultrasonography and MRI. All patients underwent excision and the tumor size was measured on histology. X-ray mammography, ultrasonography and MRI diagnostic equipment were same with the first part. The tumor size was measured on display of mammography, ultrasonography and MRI. The maximum diameter of the lesion in any plane was recorded as the tumor size. Statistical analysis was performed using an SPSS 13.0 statistical software package. Linear correlation analysis was performed and scatter plots were generated for each of the imaging measurements versus the pathology measurements. Data were analyzed using k independent samples nonparametric tests. p values smaller than 0.05 were considered significant in all cases.Results: The pathological tumor size of the 46 malignant breast tumors ranged from 5 to 70mm and the mean was 23.8mm. The mean tumor size in mammography, ultrasonography and MRI were 30.9 mm, 22.5 mm and 23.8mm, respectively. For mammographically determined size (versus pathologic size) correlation r, was 0.51 and for US it was 0.71 and for MRI it was 0.84. The difference of mammographically determined size in different lesions were statistically significant (P=0.001).Conclusion: When comparing mammography, ultrasonography and MRI for the prediction value of the pathological size of a malignant breast tumor, MRI is the best predictor, mammography is the worst predictor, US is more accurate than mammography in assessing breast cancer size. For mammography, the accuary of assessing tumor size between different lesions is different. But the measurements of mammography, ultrasonography and MRI all existed cases of underestimation or overestimation for tumor size.