Objective: To evaluate whether contrast-enhanced cone-beam breast CT (CE-CBBCT) features can risk-stratify prognostic stage in breast cancer. Methods: Overall, 168 biopsy-proven breast cancer patients were analysed: 115 patients in the training set underwent scanning using v. 1.5 CE-CBBCT between August 2019 and December 2019, whereas 53 patients in the test set underwent scanning using v. 1.0 CE-CBBCT between May 2012 and August 2014. All patients were restaged according to the American Joint Committee on Cancer eighth edition prognostic staging system. Following the combination of CE-CBBCT imaging parameters and clinicopathological factors, predictors that were correlated with stratification of prognostic stage via logistic regression were analysed. Predictive performance was assessed according to the area under the receiver operating characteristic curve (AUC). Goodness-of-fit of the models was assessed using the Hosmer-Lemeshow test. Results: As regards differentiation between prognostic stage (PS) I and II/III, increased tumour-to-breast volume ratio (TBR), rim enhancement pattern, and the presence of penetrating vessels were significant predictors for PS II/III disease (p < 0.05). The AUCs in the training and test sets were 0.967 [95% confidence interval (CI) 0.938–0.996; p < 0.001] and 0.896 (95% CI, 0.809–0.983; p = 0.001), respectively. Two features were selected in the training set of PS II vs III, including tumour volume [odds ratio (OR)=1.817, p = 0.019] and calcification (OR = 4.600, p = 0.040), achieving an AUC of 0.790 (95% CI, 0.636–0.944, p = 0.001). However, there was no significant difference in the test set of PS II vs III (P＞0.05). Conclusion: CE-CBBCT imaging biomarkers may provide a large amount of anatomical and radiobiological information for the pre-operative distinction of prognostic stage. Advances in knowledge: CE-CBBCT features have distinctive promise for stratification of prognostic stage in breast cancer.

In this paper, we proposed an effective and efficient approach to the classification of breast cancer microcalcifications and evaluated the mathematical model for calcification on mammography with a large medical dataset. We employed several semi-automatic segmentation algorithms to extract 51 calcification features from mammograms, including morphologic and textural features. We adopted extreme gradient boosting (XGBoost) to classify microcalcifications. Then, we compared other machine learning techniques, including k-nearest neighbor (kNN), adaboostM1, decision tree, random decision forest (RDF), and gradient boosting decision tree (GBDT), with XGBoost. XGBoost showed the highest accuracy (90.24%) for classifying microcalcifications, and kNN demonstrated the lowest accuracy. This result demonstrates that it is essential for the classification of microcalcification to use the feature engineering method for the selection of the best composition of features. One of the contributions of this study is to present the best composition of features for efficient classification of breast cancers. This paper finds a way to select the best discriminative features as a collection to improve the accuracy. This study showed the highest accuracy (90.24%) for classifying microcalcifications with AUC = 0.89. Moreover, we highlighted the performance of various features from the dataset and found ideal parameters for classifying microcalcifications. Furthermore, we found that the XGBoost model is suitable both in theory and practice for the classification of calcifications on mammography.

Mammography (MG) is highly sensitive for detecting microcalcifications, but has low specificity. This study investigates whether establishing a preoperative nomogram including ultrasonographic findings can help predict the likelihood of malignancy in patients with mammographic microcalcification. Between May 2012 and January 2017, 475 patients with suspicious microcalcifications detected on MG underwent ultrasonography (US). The χ2 test was used to screen risk factors among the variables. Then, a multivariate logistic regression analysis was performed to identify independent predictors of malignant microcalcifications. A mammographic nomogram (M nomogram) and mammographic-ultrasonographic nomogram (M-U nomogram) were established based on multivariate logistic regression models. The discriminatory ability and clinical utility of both nomograms were compared by the receiver operating characteristics curve and decision curve analysis. The calibration ability was evaluated using a calibration curve. Among the cases, 68.2% (324/475) were pathologically diagnosed as breast cancer and 31.8% (151/475) were benign lesions. Based on multivariate logistic regression analysis, age, clinical manifestation, morphology and distribution of microcalcifications on MG and lesions associated with microcalcifications on US were confirmed as independent predictors of malignant microcalcifications. In terms of discrimination ability, the C-index of the M-U nomogram was significantly higher than that of the M nomogram (0.917 vs 0.897, p = 0.006). The bias-corrected curve was close to the ideal line in the calibration curve. Decision curve analysis suggested that the M-U nomogram was superior to M nomogram. Combining mammographic parameters with ultrasonographic findings in a nomogram provided better performance than an M nomogram alone, especially for dense breasts, which suggests the value of ultrasonographic finding for individualized prediction of malignancy in patients with microcalcifications.

Objective . Due to the blurry edges and uneven shape of breast tumors, breast tumor segmentation can be a challenging task. Recently, deep convolution networks based approaches achieve satisfying segmentation results. However, the learned shape information of breast tumors might be lost owing to the successive convolution and down-sampling operations, resulting in limited performance. Approach . To this end, we propose a novel shape-guided segmentation (SGS) framework that guides the segmentation networks to be shape-sensitive to breast tumors by prior shape information. Different from usual segmentation networks, we guide the networks to model shape-shared representation with the assumption that shape information of breast tumors can be shared among samples. Specifically, on the one hand, we propose a shape guiding block (SGB) to provide shape guidance through a superpixel pooling-unpooling operation and attention mechanism. On the other hand, we further introduce a shared classification layer (SCL) to avoid feature inconsistency and additional computational costs. As a result, the proposed SGB and SCL can be effortlessly incorporated into mainstream segmentation networks (e.g. UNet) to compose the SGS, facilitating compact shape-friendly representation learning. Main results . Experiments conducted on a private dataset and a public dataset demonstrate the effectiveness of the SGS compared to other advanced methods. Significance . We propose a united framework to encourage existing segmentation networks to improve breast tumor segmentation by prior shape information. The source code will be made available at https://github.com/TxLin7/Shape-Seg .