Classification of MR-Detected Additional Lesions in Patients With Breast Cancer Using a Combination of Radiomics Analysis and Machine Learning

ObjectiveThis study was conducted in order to investigate the feasibility of using radiomics analysis (RA) with machine learning algorithms based on breast magnetic resonance (MR) images for discriminating malignant from benign MR-detected additional lesions in patients with primary breast cancer.Materials and MethodsOne hundred seventy-four MR-detected additional lesions (benign, n = 86; malignancy, n = 88) from 158 patients with ipsilateral primary breast cancer from a tertiary medical center were included in this retrospective study. The entire data were randomly split to training (80%) and independent test sets (20%). In addition, 25 patients (benign, n = 21; malignancy, n = 15) from another tertiary medical center were included for the external test. Radiomics features that were extracted from three regions-of-interest (ROIs; intratumor, peritumor, combined) using fat-saturated T1-weighted images obtained by subtracting pre- from postcontrast images (SUB) and T2-weighted image (T2) were utilized to train the support vector machine for the binary classification. A decision tree method was utilized to build a classifier model using clinical imaging interpretation (CII) features assessed by radiologists. Area under the receiver operating characteristic curve (AUROC), accuracy, sensitivity, and specificity were used to compare the diagnostic performance.ResultsThe RA models trained using radiomics features from the intratumor-ROI showed comparable performance to the CII model (accuracy, AUROC: 73.3%, 69.6% for the SUB RA model; 70.0%, 75.1% for the T2 RA model; 73.3%, 72.0% for the CII model). The diagnostic performance increased when the radiomics and CII features were combined to build a fusion model. The fusion model that combines the CII features and radiomics features from multiparametric MRI data demonstrated the highest performance with an accuracy of 86.7% and an AUROC of 91.1%. The external test showed a similar pattern where the fusion models demonstrated higher levels of performance compared with the RA- or CII-only models. The accuracy and AUROC of the SUB+T2 RA+CII model in the external test were 80.6% and 91.4%, respectively.ConclusionOur study demonstrated the feasibility of using RA with machine learning approach based on multiparametric MRI for quantitatively characterizing MR-detected additional lesions. The fusion model demonstrated an improved diagnostic performance over the models trained with either RA or CII alone.

Prospective comparison of CT and 18F-FDG PET/MRI in N and M staging of primary breast cancer patients: Initial results

Objectives To compare the diagnostic accuracy of contrast-enhanced thoraco-abdominal computed tomography and whole-body 18F-FDG PET/MRI in N and M staging in newly diagnosed, histopathological proven breast cancer. Material and methods A total of 80 consecutive women with newly diagnosed and histopathologically confirmed breast cancer were enrolled in this prospective study. Following inclusion criteria had to be fulfilled: (1) newly diagnosed, treatment-naive T2-tumor or higher T-stage or (2) newly diagnosed, treatment-naive triple-negative tumor of every size or (3) newly diagnosed, treatment-naive tumor with molecular high risk (T1c, Ki67 >14%, HER2neu over-expression, G3). All patients underwent a thoraco-abdominal ceCT and a whole-body 18F-FDG PET/MRI. All datasets were evaluated by two experienced radiologists in hybrid imaging regarding suspect lesion count, localization, categorization and diagnostic confidence. Images were interpreted in random order with a reading gap of at least 4 weeks to avoid recognition bias. Histopathological results as well as follow-up imaging served as reference standard. Differences in staging accuracy were assessed using Mc Nemars chi2 test. Results CT rated the N stage correctly in 64 of 80 (80%, 95% CI:70.0–87.3) patients with a sensitivity of 61.5% (CI:45.9–75.1), a specificity of 97.6% (CI:87.4–99.6), a PPV of 96% (CI:80.5–99.3), and a NPV of 72.7% (CI:59.8–82.7). Compared to this, 18F-FDG PET/MRI determined the N stage correctly in 71 of 80 (88.75%, CI:80.0–94.0) patients with a sensitivity of 82.1% (CI:67.3–91.0), a specificity of 95.1% (CI:83.9–98.7), a PPV of 94.1% (CI:80.9–98.4) and a NPV of 84.8% (CI:71.8–92.4). Differences in sensitivities were statistically significant (difference 20.6%, CI:-0.02–40.9; p = 0.008). Distant metastases were present in 7/80 patients (8.75%). 18 F-FDG PET/MRI detected all of the histopathological proven metastases without any false-positive findings, while 3 patients with bone metastases were missed in CT (sensitivity 57.1%, specificity 95.9%). Additionally, CT presented false-positive findings in 3 patients. Conclusion 18F-FDG PET/MRI has a high diagnostic potential and outperforms CT in assessing the N and M stage in patients with primary breast cancer.

Imaging Surveillance of the Reconstructed Breast in a Subset of Patients May Aid in Early Detection of Breast Cancer Recurrence

Objectives: The purpose of this study is to determine the biological markers more frequently associated with recurrence in the reconstructed breast, to evaluate the detection method, and to correlate recurrent breast cancers with the detection method. Material and Methods: An institutional review board-approved retrospective study was conducted at a single institution on 131 patients treated with mastectomy for primary breast cancer followed by breast reconstruction between 2005 and 2012. Imaging features were correlated with clinical and pathologic findings. Results: Of the 131 patients who met our inclusion criteria, 40 patients presented with breast cancer recurrence. The most common histopathologic type of primary breast cancer was invasive ductal carcinoma in 82.5% (33/40) of patients. Triple-negative breast cancer was the most common biological marker with 42.1% (16/38) of cases. Clinically, 70% (28/40) of the recurrences presented as palpable abnormalities. Of nine patients who underwent mammography, a mass was seen in eight patients. Of the 35 patients who underwent ultrasound evaluation, an irregular mass was found in 48.6% (17/35) of patients. Nine patients with recurrent breast cancer underwent breast MRI, and MRI showed an irregular enhancing mass in four patients, an oval mass in four patients, and skin and trabecular thickening in one patient. About 55% of patients with recurrent breast cancer were found to have distant metastases. Conclusion: Patients at higher risk for locoregional recurrence may benefit from imaging surveillance in order to detect early local recurrences.