Background: The American College of Radiology Breast Imaging Reporting and Data System (ACR BI-RADS) used with ultrasonography cannot guide the individual management of solid breast tumors, but preoperative core biopsy categories (CBCs) can. We aimed to use machine learning to analyze clinical and ultrasonic features for predicting CBCs and to aid in the development of a new ultrasound (US) imaging reporting system for solid tumors of the breast.
Methods: This retrospective study included women with solid breast tumors who underwent US-guided core needle biopsy from March 1, 2019, to December 31, 2019. All patients were randomly assigned to a training or validation cohort (7:3 ratio). CBC was predicted using 5 machine learning models: random forest (RF), support vector machine (SVM), k-nearest-neighbor (KNN), multilayer perceptron (MLP), and ridge regression (RR). In the validation cohort, the area under the curve (AUC) and accuracy were ascertained for every algorithm. Based on AUC values, the optimal algorithm was determined, and the features' importance was depicted.
Results: A total of 1,082 female patients were included (age range, 12-96 years; mean age ± standard deviation, 42.22±13.37 years). The proportion of the 4 CBCs was 4% (44/1,185) for the B1 group, 60% (714/1,185) for the B2 group, 5% (57/1,185) for the B3 group, and 31% (370/1,185) for the B5 group. In the validation cohort, AUCs of the optimal algorithm constructed RF were 0.78, 0.88, 0.64, and 0.92 for B1, B2, B3, and B5, respectively, with an accuracy of 0.82.
Conclusions: Machine learning could strongly predict CBC, particularly in B2 and B5 categories of solid breast tumors, with RF being the optimal machine learning model.
Keywords: Ultrasound (US); core biopsy categories (CBCs); machine learning; prediction; solid breast tumor.
2023 Quantitative Imaging in Medicine and Surgery. All rights reserved.