Application Of Diffusion Weighted Imaging With Various Mathematical Models In Identifying Benign And Malignant Breast Lesions

Background and objectives:At present, breast cancer is the highest incidence of cancer in China. It is the sixth cause of death in Chinese women after lung cancer, stomach cancer, liver cancer, esophageal cancer and colorectal cancer. Both reproductive and hormonal factors, obesity and low levels of physical activity in postmenopausal women were considered to be the risk factors for breast cancer, and the age of onset of breast cancer was constantly younger. So early diagnosis and differential diagnosis of benign and malignant breast tumors are very important for the development of appropriate treatment options, appropriate surgical procedures and prognosis assessment. Breast ultrasound is one of the important methods for early detection of breast cancer. Breast X-ray examination has radiation, and there is a diagnostic defect to the dense breast. MRI has good soft tissue resolution, no radiation, its value for the diagnosis of breast disease has been fully affirmed. Especially for traditional ultrasound and X-ray examination to identify atypical lesions, which can analysis from lesion morphology, signal strength, dynamic enhanced hemodynamics and functional magnetic resonance applications, found in small lesions had a significant advantage. But there are some limitations in the diagnosis and differential diagnosis of breast lesions by routine MRI examination. Diffusion weighted imaging (DWI) by the movement state detection to the organization of internal water molecules to reflect the structural characteristics of the lesions and provide valuable information for the differential diagnosis of the lesions.Magnetic resonance diffusion weighted imaging using random displacement of water molecules in living organisms (i.e., Brown motion) for noninvasive imaging. Because of the high sensitivity of DWI to the cell density, the relative content of tumor cells, water content and perfusion, it has been more and more widely used in the evaluation of breast lesions. Always use single B value DWI single index model (mono-eXponential model) and apparent diffusion coefficient apparent diffusion coefficient ADC) value analysis of breast tumor, but due to the single index model of ADC value integrate the information of living tissue in real diffusion and pseudo perfusion, which reflected the organization of diffusion characteristics influenced by the capillary microcirculation. Recently proposed multi b value DWI double exponential decay model, is plain intra-voxel incoherent motion(IVIM). The real biological tissue water molecular diffusion is separated from microcirculation perfusion in capillary, also reflect biological tissue cell composition and perfusion information, the measurement data are more close to the biological tissue diffusion values. Another mathematical model of multi b value DWI is stretched exponential model, by describing the distribution of internal water molecules diffusion index (distributed diffusion coefficient, DDC) and diffusion heterogeneity index (heterogeneity index of intravoxel diffusion, alpha), reflecting more information of the organization biological features.This topic research analyzed the magnetic resonance diffusion weighted imaging of various mathematical models in benign and malignant tumors of the breast, discuss the effect of DWI of the single index model, double exponential model and stretched exponential model in identifying benign and malignant breast tumors.Materials and methods:87 cases (92 lesions) were collected by imaging examination and pathology confirmed. Patients’ age were 20-70 years old. All patients were examined by Discovery MR7503.0T MR GE, followed by routine MRI scan, dynamic enhancement, single DWI value B and multi DWI B scan. Single B value DWI using spin echo-echo planner imaging (SE-EPI) sequence, axial scanning, b value of 0,800 s/mm2. The corresponding apparent diffusion coefficients (diffusion coefficient, ADC) are obtained at the same time, and the corresponding apparent diffusion coefficients (apparent) are obtained. B value DWI selected by 0-3000 s/mm212 different b values by axial scan (0, 10,20,30,50,100,200,400,800 and 1200,2000,3000 s/mm2), after the processing software were carried out on the calculation of the biexponential model and stretched-exponential model.Biexponential model to obtain the parameters:the standard apparent diffusion coefficient(ADCst), slow apparent diffusion coefficient (ADCslow), fast apparent diffusion coefficient(ADCfast), the proportion of fast diffusion component(ffast). Distributed diffusion coefficient (DDC) and heterogeneity index of intravoxel diffusion are generated by the stretched-exponential model. The value of the region of interest (ROI) in the real area of benign and malignant breast tumors was compared with the value of each parameter in the DWI index model. In order to quantitatively analyze the diffusion of water molecules in patients with normal breast as compared.All statistical analysis using SPSS 19.0 software, first by non parametric one sample Kolmogorov Smirnov test of ADC, ADCst ADCslow ADCfast, ffast, DDC, alpha values for normal distribution test, ADC, ADCst ADCslow ADCfast, ffast, DDC, alpha values are in accordance with the gaussian distribution. Measurement data with mean ±standard deviation (x±s), ADC, ADCst, ADCslow, ADCfast, ffast, DDC, alpha values from benign and malignant group compared using independent samples t test, P <0.05 was considered statistically significant. The differences of ADC, ADCst, ADCslow, and DDC values between benign and malignant groups were analyzed by paired t test, respectively, P<0.05 was considered statistically significant. Draw the ROC curve of ADC, ADCst, ADCslow, ADCfast, ffast, DDC and alpha values, comparison of the area under the curve (AUC). The best diagnostic threshold and the corresponding sensitivity and specificity were obtained according to the largest Youden index. The relationships between ADC, ADCst, ADCslow, DDC and were analyzed by Pearson, P<0.05 was considered statistically significant.Receiver operator characteristic (ROC) curve was used to evaluate the diagnostic efficacy and the optimal diagnostic threshold value of the single exponential model, the biexponential model and stretched-exponential model. Results:1. The ADC of breast lesions in decreasing order is cysts, normal glands, benign tumor, malignant tumor, mastitis.2. The ADC and ADCst, ADCslow and ffast values of cysts were all larger than normal glandular tissues, ADC, ADCst and ffast had statistical differences, and there was no significant difference on ADCslow between the two groups. The ADCfast value was lower than that of the normal breast tissue, and there was no significant difference between the two groups. The DDC and alpha values of the cysts were larger than those of normal breast tissue, and the difference was statistically significant.3. The ADC and ADCst ADCslow ffast values of mastitis were less than normal glandular tissue, and there was a significant difference. ADCfast of mastitis was higher than that in normal breast tissue, but no significant difference. The DDC and alpha of mastitis were less than the normal breast tissue, and the difference was statistically significant.4. The ADC value of the benign lesions was lower than that of the normal glandular tissue, and the difference was statistically significant. The ADCst and ADCslow of the benign lesions were all smaller than the normal glandular tissues, but there was no statistical difference. The ffast value was significantly smaller than that of normal glandular tissue, and the difference was statistically significant. The ADCfast value of benign lesions was higher than that in normal breast tissue, but there was no significant difference. The DDC and alpha value of the benign lesions were all smaller than that of the normal breast tissue, and the difference was statistically significant.5. The ADC and ADCst, ADCslow and ffast value of breast malignant lesions were all less than normal glandular tissues, and there were significant differences. The ADCfast value of malignant breast lesions was higher than that in normal breast tissue, and the difference was statistically significant. The DDC and alpha value of malignant breast lesions were all smaller than normal breast tissue, and the difference was statistically significant.6. The ADC and ADCst, ADCslow, ffast values of breast benign lesions were higher than that of breast malignant lesions, and there were significant differences. ADCfast value of benign breast lesions was lower than that of malignant breast lesions, and the difference was statistically significant. The DDC and alpha value of benign breast lesions were all larger than that of malignant breast lesions, and the difference was statistically significant.7. In benign and malignant breast tumors, the ADC value was significantly higher than ADCslow value(P<.001). The DDC value is larger than ADCslow, ADCst (P<0.001).8. Pearson correlation coefficients between ADC and DDC in benign and malignant breast tumors were:r=0.762 (P<0.001) and r=0.785 (P<0.001). ADCst and DDC were:r=0.829 (P<0.001) and r=0.938 (P<0.001), was significantly correlated. ADCslow and DDC were r=0.528 (P=0.001) and r=0.764 (P<0.001), respectively. There was no significant correlation between the alpha and DDC (P>0.05).9. In identify of benign and malignant breast tumors, the ADC,ADCslow and DDC area under the ROC curve were 0.923,0.920 and 0.903, respectively. According to the Youden index, the sensitivity of ADCslow to identify benign and malignant breast lesions was 93.9%, and the specificity of ADCst was 95%. Conclusion:1. DWI with single index model, biexponential model and stretched-exponential model can provide an important basis for the quantitative identification of benign and malignant breast tumors.2. ADC and ADCslow value has a high diagnostic value, DDC has the important value with potential.3. In the differentiation of benign and malignant breast tumors, the double exponential model and the stretched exponential model DWI are more refined than the traditional single index model DWI.