• breast calcifications
• diagnosis.
• digital breast tomosynthesis
• radiomic

According to the Breast Imaging Reporting and Data System (BI-RADS) microcalcifications category 4 includes those findings that have from 2% to 95% chance to be a neoplasia. Regarding the detection of the microcalcifications, mammography is a test with high sensibility (about 95%) but low specificity (about 41%) with a positive predictive value (PPV) inferior to 30%. To date the definitive diagnosis of suspicious microcalcifications detected on imaging necessarily requires a histopathologic examination and, therefore, a biopsy sample. Considering that the majority of the suspicious microcalcifications addressed to biopsy (70-80%) turn out to be benign at the histopathological examination, it is clear that many of these biopsies could be avoided. An instrument capable of reducing the rate of false-positive findings is needed. This hypothetical tool should overcome the large use of biopsy which is an invasive, not risk-free, and expensive technique. Radiomics is a new re-search tool well suited to this scenario. Indeed, radiomics represents an advanced ana-lytic methodology that uses and studies the quantitative features extracted from bio-medical images to generate imaging biomarkers for better support in the clinical management of the patient. The study aims to investigate whether quantitative radiomic features extracted from digital breast tomosynthesis (DBT) can be an additional and useful tool to discriminate between benign and malignant BI-RADS category 4 microcalcification.
This retrospective study included 252 female patients with BI-RADS category 4 microcalcifications addressed to vacuum-assisted breast biopsy (VABB) in our center. According to micro-histopathology, the patients were divided into two groups: 126 patients with benign lesions and 126 patients with certain or possible malignancies. A total of 91 radiomic features were extracted from the centering tomosynthesis of each patient and the 12 most representative features were selected by using the agglomerative hierarchical clustering method. The binary classification task of the two groups was carried out by means of four different machine-learning algorithms (i.e., linear support vector machine (SVM), radial basis function (RBF) SVM, logistic regression (LR), and random forest (RF)). Accuracy, sensitivity, sensibility, and the area under the curve (AUC) were calculated for each of them.
The best performance was achieved using the RF classifier (AUC=0.59, 95% confidence interval 0.57–0.60; sensitivity=0.56, 95% CI 0.54–0.58; specificity=0.61, 95% CI 0.59–0.63; accuracy=0.58, 95% CI 0.57–0.59).
Our study has some limitations: it was a retrospective single-center design; although we had not a small data set, it could be amplified in future studies; the clinical risk factors were not incorporated; we did not correlate the radiomic results with breast density which can influence the risk to develop breast cancer; for the manual segmentation of the process a single radiologist was involved, therefore, was not possible to evaluate the reliability of the intra- and inter-observer processes.
In conclusion, DBT-based radiomic analysis seems to show only a limited potential in discriminating benign from malignant microcalcifications. Therefore, radiomic features alone are not able to define the clinical management of patients with BI-RADS category 4 microcalcifications. However, our results did not exclude that a further improved classification model can reduce the false-positive rate and adjust the radiologic cut-off for image-guided breast biopsy. We believe that with further large-scale studies capable of overcoming the limits of our work it might be possible to obtain a radiomic classification model as a supplement to the BI-RADS for a better selection of patients with suspicious microcalcifications that need VABB.