Evaluation of diffusion kurtosis imaging in the differential diagnosis of breast lesions / 中华放射学杂志

ABSTRACT

Objective To evaluate the role of the diffusion kurtosis imaging(DKI)in the differential diagnosis of breast lesions. Methods Seventy five breast lesions(32 benign and 43 malignant)in 72 patients confirmed by histopathology were studied. All patients underwent 3.0 T MR examinations, including T1WI, T2WI, T2WI-spectral adiabatic inversion recovery, 4b diffusion-weighted imaging, and dynamic contrast-enhanced MR imaging(DCE-MRI). Data were post-processed by mono-exponential and diffusion kurtosis models for quantitation of ADC, apparent diffusion for non-Gaussian distribution(D), and apparent kurtosis coefficient(K). All breast lesions were described with the classification by breast imaging report and data system(BI-RADS). Lesions with BI-RADS class 4B or above were rated as malignancy. Independent sample t test was used to compare the ADC, D, and K value differences between benign and malignant lesions . ROC analysis was performed to assess the role of ADC, D, K value, and BI-RADS in the differential diagnosis of breast lesions. The morphological characteristics, time-signal curve(TIC)type, and other differences between benign and malignant lesions were analyzed with Chi-square test. Results ADC and D values were significantly lower in malignant than in benign lesions(P<0.01). Conversely, K value was significantly higher in malignant lesions than in benign ones(P<0.01). The shape of the benign and malignant breast lesions, edge, enhancement mode, TIC, and BI-RADS classification difference had statistical significance(P<0.05, respectively). The areas under the ROC curve of ADC, D, K, DCE-MRI, and DCE-MRI combined with K value were 0.857, 0.884, 0.949, 0.806, and 0.958, respectively. DCE-MRI combined with K value had the highest diagnosis efficiency. At a cutoff value of K= 0.856, the sensitivity and specificity were 83.7% and 93.8%, respectively. Conclusions DKI model showed higher diagnostic efficiency than that of traditional DWI model. DCE-MRI combined with K value can increase the diagnostic efficiency in breast lesions.