Radiomics Analysis on Digital Breast Tomosynthesis: Preoperative Evaluation of Lymphovascular Invasion Status in Invasive Breast Cancer

Abstract

To develop a digital breast tomosynthesis (DBT)-based radiomics nomogram for preoperative evaluation of lymphovascular invasion (LVI) status in patients with invasive breast cancer (IBC). A total of 135 patients with pathologically confirmed IBC who underwent preoperative DBT from July 2018 to May 2020 were retrospectively enrolled and randomized into the training and validation sets. Radiomics feature extraction was performed on the volume of interest (VOI) manually outlined. A four-step algorithmic was applied to screen the features with the highest predictive power in the training set for constructing the radiomics signature and calculating the correspondent radiomics score (Rad-score). Logistic regression analyses were utilized to develop a combined radiomics model that incorporated the DBT-reported clinicoradiological semantic features and Rad-score, which was visualized as a radiomics nomogram. The percentage of LVI-positive patients was 60.2% and 59.5% in the training and validation sets, respectively. The radiomics signature was constructed based on nine features selected from the 1218 radiomics features extracted. Higher Rad-score, maximum tumor diameter, and spiculate margin were independent risk factors for LVI. The area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, and specificity of the radiomics nomogram were 0.905, 72.7%, and 94.6% in the training set, and 0.835, 80.0%, and 76.5% in the validation set, respectively; this data was higher than models incorporating clinicoradiological semantic features alone or the radiomics signature in both sets. Preoperative DBT-based combined radiomic nomogram could be a potential biomarker for LVI in patients with IBC.