Nomogram model based on multiparametric MRI combined with clinical features in identifying benign and malignant BI-RADS 4 lesions / 中华放射学杂志

ABSTRACT

Objective:To investigate the efficacy of the nomogram model based on multiparametric MRI combined with clinical features for differential diagnosis of benign and malignant breast imaging reporting and data system (BI-RADS) 4 lesions.Methods:This study was a cross-sectional study. Clinical and imaging data of 56 patients (66 lesions) with pathologically confirmed BI-RADS 4 breast lesions from January 2020 to June 2022 at Second Affiliated Hospital of Shantou University Medical College were retrospectively analyzed. The patients were all females aged 42 (17, 71) years old. All patients underwent the breast MRI, including T 1WI, T 2WI, diffusion-weighted imaging, diffusion kurtosis imaging (DKI), and dynamic-enhanced MRI (DCE-MRI), and the patient clinical characteristics, imaging characteristics as well as relevant MRI quantitative parameters were recorded. Comparisons of the indicators of benign and malignant BI-RADS 4 lesions were performed by sample t-test , Mann-Whitney U, or χ 2 test. The least absolute shrinkage and selection operator regression was utilized to further select indicators with statistically significant differences in univariate analyses, and finally, nomogram models were constructed and reclassified all the lesions. Results:Of the 66 lesions in 56 patients, 24 lesions were found in 24 malignant patients and 42 lesions in 32 benign patients. The differences in age, body mass index, and menopausal status between benign and malignant patients were statistically significant (all P<0.05); the differences in tumor longest diameter, type of lesion enhancement, time-single intensity curve type, mean diffusivity and mean kurtosis (MK) between benign and malignant lesions were statistically significant (all P<0.05). After feature selection, MK ( OR=27.952, 95% CI 1.301-600.348, P=0.033), age ( OR=1.140, 95%CI 1.040-1.249, P=0.005), and the type of lesion enhancement ( OR=0.045, 95%CI 0.006-0.316, P=0.005) were the independent influences in predicting BI-RADS 4 malignant lesions. Using this to construct a nomogram model, its area under the curve for predicting BI-RADS 4 malignant lesions was 0.946, and the accuracy of reclassifying 66 BI-RADS 4 lesions as benign versus malignant was 86.36% (57/66). Conclusion:The nomogram model constructed with MK from DKI parameters, the type of lesion enhancement from DCE-MRI, and age is valuable in diagnosing the benign and malignant nature of BI-RADS 4 lesions.