We aimed to investigate the efficacy of computer-aided detection (CAD) for MRI in the assessment of tumor extent, lymph node status, and multifocality in invasive breast cancers in comparison with other breast imaging modalities. Two radiologists measured the maximum tumor size, as well as, analyzed lymph node status and multifocality in 86 patients with invasive breast cancers using mammography, ultrasound, CT, MRI with and without CAD, and 18-fludeoxyglucose positron emission tomography (FDG-PET). The assessed data were compared with pathology. For tumor extent, there were no significant differences between pathological size and measured size using mammography, ultrasound, CT, or MRI with and without CAD (P > 0.05). For evaluation of lymph node status, ultrasound had the best kappa coefficients (0.522) for agreement between imaging and pathology, and diagnostic performance with 92.1% specificity and 90.0% positive predictive value. For multifocality, MRI with CAD had the highest area under the receiver operating characteristic curve (AUC = 0.888). CAD for MRI is feasible to assess tumor extent and multifocality in invasive breast cancer patients. However, CAD is not effective in evaluation of nodal status.
B-mode and vascular US features reflect genomic alterations associated with hormone receptor status, angiogenesis, or prognosis in breast cancer. Background Radiogenomic investigations for breast cancer provide an understanding of tumor heterogeneity and discover image phenotypes of genetic variation. However, there is little research on the...
This prospective study enrolled 147 women with invasive breast cancer who underwent low-dose breast CT (80 kVp, 25 mAs, 1.01–1.38 mSv) before treatment. From each tumor, we extracted eight perfusion parameters using the maximum slope algorithm and 36 texture parameters using the filtered histogram technique. Relationships between CT parameters and histological factors were analyzed using five machine learning algorithms. Performance was compared using the area under the receiver-operating characteristic curve (AUC) with the DeLong test. The AUCs of the machine learning models increased when using both features instead of the perfusion or texture features alone. The random forest model that integrated texture and perfusion features was the best model for prediction (AUC = 0.76). In the integrated random forest model, the AUCs for predicting human epidermal growth factor receptor 2 positivity, estrogen receptor positivity, progesterone receptor positivity, ki67 positivity, high tumor grade, and molecular subtype were 0.86, 0.76, 0.69, 0.65, 0.75, and 0.79, respectively. Entropy of pre- and postcontrast images and perfusion, time to peak, and peak enhancement intensity of hot spots are the five most important CT parameters for prediction. In conclusion, machine learning using texture and perfusion characteristics of breast cancer with low-dose CT has potential value for predicting prognostic factors and risk stratification in breast cancer patients.
Among artificial intelligence–missed invasive breast cancers on mammograms, the human epidermal growth factor receptor 2–enriched subtype had the lowest false-negative rate, and dense breasts, nonmammary zone locations, architectural distortions, and amorphous microcalcifications were reasons for misses.