A nomogram based on multiparametric magnetic resonance imaging improves the diagnostic performance of breast lesions diagnosed as BI-RADS category 4: A comparative study with the Kaiser score

Xiao Yang; Zhou Lu; Xiaoying Tan; Lin Shao; Jie Shi; Weiqiang Dou; Zongqiong Sun

doi:10.1016/j.ejrad.2025.111920

A nomogram based on multiparametric magnetic resonance imaging improves the diagnostic performance of breast lesions diagnosed as BI-RADS category 4: A comparative study with the Kaiser score

Eur J Radiol. 2025 Jan 3:183:111920. doi: 10.1016/j.ejrad.2025.111920. Online ahead of print.

Authors

Xiao Yang¹, Zhou Lu¹, Xiaoying Tan¹, Lin Shao¹, Jie Shi², Weiqiang Dou², Zongqiong Sun³

Affiliations

¹ Department of Radiology, Affiliated Hospital of Jiangnan University, Wuxi City, Jiangsu Province 214062, China.
² GE Healthcare, MR Research China, Beijing 100176, China.
³ Department of Radiology, Affiliated Hospital of Jiangnan University, Wuxi City, Jiangsu Province 214062, China. Electronic address: qiong953780@163.com.

PMID: 39793481
DOI: 10.1016/j.ejrad.2025.111920

Abstract

Purpose: To construct a nomogram combining Kaiser score (KS), synthetic MRI (syMRI) parameters, apparent diffusion coefficient (ADC), and clinical features to distinguish benign and malignant breast lesions better.

Methods: From December 2022 to February 2024, a retrospective cohort of 168 patients with breast lesions diagnosed as Breast Imaging Reporting and Data System (BI-RADS) category 4 by ultrasound and/or mammography was included. The research population was divided into the training set (n = 117) and the validation set (n = 51) by random sampling with a ratio of 7:3. Breast lesions' KS, ADC, relaxation time of syMRI, and clinical and imaging features were statistically analyzed and compared between malignant and benign groups. Two experienced radiologists independently assigned KS, and measured quantitative values of ADC and parameters of syMRI, and the intraclass correlation coefficient (ICC) was calculated. Independent predictors were identified by univariable and multivariable logistic regression analysis. Then, a nomogram was established, and its performance was evaluated by the area under the curve (AUC), calibration curve, and decision curve.

Results: There were 168 lesions (118 malignant and 50 benign) in 168 female patients confirmed by histopathology. The interobserver agreement for each quantitative parameter was excellent. Older patient (OR = 1.091, 95 % confidence interval [CI]: 1.017-1.170, P = 0.014), higher lesions' KS (OR = 288.431, 95 % CI: 34.930-2381.654, P < 0.001), lower ADC (OR = 0.077, 95 % CI: 0.011-0.558, P = 0.011), and lower T2 relaxation time (OR = 0.918, 95 % CI: 0.868-0.972, P = 0.003) were independent predictors of breast malignancies and utilized to establish the nomogram. The accuracy of KS, ADC, T2, and patient age in predicting malignant breast lesions was 88.89 %, 79.48 %, 82.05 %, and 58.97 %, respectively. No significant differences in AUCs of KS, ADC and T2 were observed in distinguishing benign from malignant breast lesions. The nomogram yielded higher AUCs of 0.968 (0.934-0.996) and 0.959 (0.863-0.995) in training and validation sets than KS, ADC, T2, and patient age (p < 0.05).

Conclusion: Although there were no significant differences among the AUCs of KS, ADC, and T2, the constructed nomogram incorporating these parameters significantly improves diagnostic performance for distinguishing benign and malignant BI-RADS 4 breast lesions. Future external validation is needed in practical applications.

Keywords: Breast lesion; Diagnosis; Kaiser score; Nomogram; Synthetic magnetic resonance imaging.