Histogram analysis of diffusion kurtosis imaging in the differentiation of malignant from benign breast lesions

Objective: To assess the diagnostic accuracy of histogram analysis of diffusion kurtosis imaging (DKI) in breast lesions. Materials and methods: Our institutional review board approved this retrospective study. Seventy-two breast lesions (30 benign and 42 malignant) in 71 patients were histopathologically confirmed. All breast lesions were evaluated by 3.0-T diffusion-weighted imaging (DWI) with 4 b-values of 0, 500, 800, and 2000s/mm(2) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). Histogram analyses of conventional DWI and DKI were performed using FireVoxel software for whole lesions. The parameters included apparent diffusion coefficient (ADC), diffusivity (D), and kurtosis (K). The metrics of ADC and DKI parameters (D and K) for benign lesions were compared with those for malignant lesions. The effectiveness of the ADC and DKI parameters (D and K) for diagnosing breast lesions was analysed using receiver operating characteristic (ROC) regression models. Results: For the malignant breast lesions, the mean, median, and 10th/25th/75th percentile values of ADC and D were significantly lower, while the skewness of ADC and D were significantly higher in comparison of the benign lesions(all p < 0.05). The malignant lesions had significantly higher mean, median, and 10th/25th/75th/90th percentile K values than did the benign lesions (all p < 0.05). Within each set of parameters, the 10th percentile ADC (Az = 0.752) and D, (Az = 0.834) coupled with the 75th percentile K (Az = 0.904) were the best metrics for differentiating benign from malignant breast lesions. After comparing the parameters in pairs, the Az for the 75th percentile K was significantly higher than that for the 10th percentile ADC (p = 0.0321) in differentiating benign from malignant breast lesions. When comparing the combination of the 75th percentile K and the 10th percentile D (Az = 0.937) with the 10th percentile D, 75th percentile K and the mean K, a significantly higher Az was observed for the combination than that for the 10th percentile D and the mean K (p = 0.0097 and p = 0.0431, respectively). The diagnostic sensitivity and specificity of the combination of the 75th percentile K and the 10th percentile D were 85.71% and 93.33%, respectively. Conclusion: Histogram analysis of DKI can accurately reflect the histologic characteristics and heterogeneity and is a reliable method for diagnosing breast lesions.