Anisotropic parameters of diffusion tensor imaging in breast: a preliminary study for detection and differentiation of malignant tumors

Objective: To investigate the diagnostic value of DTI anisotropy parameters in breast malignant tumors. Methods: Fifty four patients,including 33 patients with malignant tumors and 21 patients with benign lesions,were retrospectively analyzed. The E1, MD and FA of lesions were measured and compared by paired t test between the malignant tumors and the contralateral healthy breast tissue. The difference between malignant tumors and benign lesions was analyzed by independent sample t test. Logistic regression analysis was made using E1, FA, MD as predictors in detecting and differentiating the malignant tumors, ROC curve analysis was performed to compare diagnostic performance based on the area under the curve (AUC). Results: E1, MD and FA in malignant tumors were (0.99 ± 0.12) × 10^-3 mm²/s,(0.85 ±0.26) × 10^-3 mm²/s and 0.20 ±0.08 respectively,and those in normal breast tissues were(1.46 ± 0.55) × 10^-3 mm²/s, (1.48 ± 0.44) × 10^-3 mm²/s and 0.29 ± 0.17 respectively.Those parameters in benign lesions were (1.80 ±0.42) × 10^-3 mm²/s, (1.38 ±0.52) × 10^-3 mm²/s and 0.22 ± 0.10 respectively. Significant statistic differences were found between malignant tumors and normal breast tissues in E1, MD and FA (t=-4.889, -6.449, -2.842; P<0.01). Significant statistic differences were also found between malignant tumors and benign lesions in E1 and MD (t=-10.476, -4.394; P<0.01) with no difference found in FA (P>0.05). E1, MD and FA are independent predictors in malignant tumors' detection, and the combination of E1, MD and FA significantly improved discrimination between cancer and normal tissue over each one alone with the sensitivity 97.0% (32/33), specificity 97.0% (32/33), accuracy 97.0% (64/66). Combination of E1 and MD had a similar AUC with E1 and a more AUC than MD and FA, with the sensitivity 97.0% (32/33), specificity 100.0% (21/21), accuracy 98.1% (53/54). Conclusion: The regression model combining E1, MD and FA is most valuable in breast cancer detection and E1 is the preferred index for the differentiation of breast cancers from benigin lesions. (authors)