Objective To explore histogram parameters and texture features of benign and malignant breast lesion on turbo inversion recovery magnitude(Tirm) sequence, and evaluate which parameter could help best differentiate benign from malignant breast lesion.
Methods This retrospective study included 100 breast cancer patients who underwent conventional MRI and confirmed pathologically. Texture features were derived from the gray level co-occurrence matrix(GLCM), and entropy, energy, correlation, inertia, inverse difference moment, cluster prominence and mean value, skewness, kurtosis of histogram parameters were calculated. Then we assessed the diagnosis efficacy with these parameters among the variety kinds of benign and malignant breast lesions respectively, and establish the receiver-operating characteristic curve(ROC). We assessed the differences between benign and malignant groups by the Yoden index combined with clinic for the cut-off values.
Results The differences of correlation, cluster prominence, mean value and kurtosis were statistically significant between the benign and malignant groups(all P < 0.001); the differences of energy, entropy, inertia, inverse difference moment and skewness were not statistically significant(all P > 0.05). The results of ROC with correlation, cluster prominence, mean value and kurtosis on the diagnosis of benign and malignant breast lesions were statistically significant, respectively(P < 0.001). Combined the four parameters on diagnosis benign and malignant lesions, the AUC was 0.868, the sensitivity was 90.57% and the specificity was 72.34%.
Conclusion The histogram analysis and texture analysis based on Tirm sequence could be used for the differential diagnosis of benign and malignant breast lesions. Correlation, cluster prominence, mean value and kurtosis have certain significance in the diagnosis. The combined diagnosis could improve the differential ability of benign and malignant breast lesions.
DownLoad: