Deep Learning Models for Automated Assessment of Breast Density Using Multiple Mammographic Image Types.

Recently, convolutional neural network (CNN) models have been proposed to automate the assessment of breast density, breast cancer detection or risk stratification using single image modality. However, analysis of breast density using multiple mammographic types using clinical data has not been reported in the literature. In this study, we investigate pre-trained EfficientNetB0 deep learning (DL) models for automated assessment of breast density using multiple mammographic types with and without clinical information to improve reliability and versatility of reporting. 120,000 for-processing and for-presentation full-field digital mammograms (FFDM), digital breast tomosynthesis (DBT), and synthesized 2D images from 5032 women were retrospectively analyzed. Each participant underwent up to 3 screening examinations and completed a questionnaire at each screening encounter. Pre-trained EfficientNetB0 DL models with or without clinical history were optimized. The DL models were evaluated using BI-RADS (fatty, scattered fibroglandular densities, heterogeneously dense, or extremely dense) versus binary (non-dense or dense) density classification. Pre-trained EfficientNetB0 model performances were compared using inter-observer and commercial software (Volpara) variabilities. Results show that the average Fleiss' Kappa score between-observers ranged from 0.31-0.50 and 0.55-0.69 for the BI-RADS and binary classifications, respectively, showing higher uncertainty among experts. Volpara-observer agreement was 0.33 and 0.54 for BI-RADS and binary classifications, respectively, showing fair to moderate agreement. However, our proposed pre-trained EfficientNetB0 DL models-observer agreement was 0.61-0.66 and 0.70-0.75 for BI-RADS and binary classifications, respectively, showing moderate to substantial agreement. Overall results show that the best breast density estimation was achieved using for-presentation FFDM and DBT images without added clinical information. Pre-trained EfficientNetB0 model can automatically assess breast density from any images modality type, with the best results obtained from for-presentation FFDM and DBT, which are the most common image archived in clinical practice.

Overutilization of Health Care Resources for Breast Pain.

OBJECTIVE: The objective of this study is to analyze the incidence of women with breast pain who present to an imaging center and assess the imaging findings, outcomes, and workup costs at breast imaging centers affiliated with one institution. MATERIALS AND METHODS: Demographic characteristics of and imaging findings for female patients presenting with breast pain at three community breast imaging centers between January 1, 2014, and December 31, 2014, were reviewed. Patients who were pregnant, were lactating, had a history of breast cancer, or presented with palpable nipple or skin findings were excluded. RESULTS: A total of 799 patients met the study criteria. Pain was diffuse in 30%, was focal in 30%, and was not localized in 40%. Of the 799 patients with breast pain, 790 (99%) presented for a diagnostic evaluation; 759 (95%) of these evaluated patients had negative findings. A benign sonographic correlate was detected in the area of pain in 5% of patients (39/799). One patient had a single cancer detected in the contralateral asymptomatic breast. When correlations between breast pain and the presence of cancer in the study patients were compared with the concurrent cancer detection rate in the screening population (5.5 cases per 1000 examinations performed), breast pain was not found to be a sign of breast cancer (p = 0.027). Patients younger than 40 years (316/799) underwent a total of 454 workup studies for breast pain; all findings were benign, and the cost of these studies was $87,322. Patients 40 years or older (483/799) underwent 745 workup studies, for a cost of $152,732. CONCLUSION: Breast pain represents an area of overutilization of health care resources. For female patients who present with pure breast pain, breast imaging centers should consider the following imaging protocols and education for referring physicians: an annual screening mammogram should be recommended for women 40 years or older, and reassurance without imaging should be offered to patients younger than 40 years.

Biopsy Feasibility Trial for Breast Cancer Pathologic Complete Response Detection after Neoadjuvant Chemotherapy: Imaging Assessment and Correlation Endpoints.

PURPOSE: This study was designed to present the secondary imaging endpoints of the trial for evaluating mammogram (MMG), ultrasound (US) and image guided biopsy (IGBx) assessment of pathologic complete response (pCR) in breast cancer (BC) patients undergoing neoadjuvant chemotherapy (NAC). METHODS: Patients with T1-3, N0-3, M0 triple-negative or HER2-positive BC who received NAC were enrolled in an Institutional Review Board-approved prospective, clinical trial. Patients underwent US and MMG at baseline and after NAC. Images were evaluated for residual abnormality and to determine modality for IGBx [US-guided (USG) or stereotactic guided (SG)]. Fine-needle aspiration and 9-G, vacuum-assisted core biopsy (VACBx) of tumor bed was performed after NAC and was compared with histopathology at surgery. RESULTS: Forty patients were enrolled. Median age was 50.5 (range 26-76) years; median baseline tumor size was 2.4 cm (range 0.8-6.3) and 1 cm (range 0-5.5) after NAC. Nineteen patients had pCR: 6 (32%) had residual Ca2+ presurgery, 5 (26%) residual mass, 1 (5%) mass with calcifications, and 7 (37%) no residual imaging abnormality. Sensitivity, specificity, and accuracy of US, MMG, and IGBx for pCR were 47/95/73%, 53/90/73%, and 100/95/98%, respectively. Twenty-five (63%) patients had SGBx and 15 (37%) had US-guided biopsy (USGBx). Median number of cores was higher with SGBx (12, range 6-14) than with USGBx (8, range 4-12), p < 0.002. Positive predictive value for pCR was significantly higher for SG VACBx than for USG VACBx (100 vs. 60%, p < 0.05). CONCLUSIONS: SG VACBx is the preferred IGBx modality for identifying patients with pCR for trials testing the safety of eliminating surgery.

Incremental cancer detection using breast ultrasonography versus breast magnetic resonance imaging in the evaluation of newly diagnosed breast cancer patients.

OBJECTIVE: To compare the incremental cancer detection rate (ICDR) using bilateral whole-breast ultrasonography (BWBUS) vs dynamic contrast-enhanced MRI in patients with primary breast cancer. METHODS: A retrospective database search in a single institution identified 259 patients with breast cancer diagnosed from January 2011 to August 2014 who underwent mammography, BWBUS and MRI before surgery. Patient characteristics, tumour characteristics and lesions seen on each imaging modality were recorded. The sensitivity, specificity and accuracy for each modality were calculated. ICDRs according to index tumour histology and receptor status were also evaluated. The effect of additional cancer detection on surgical planning was obtained from the medical records. RESULTS: A total of 266 additional lesions beyond 273 index malignancies were seen on at least 1 modality, of which 121 (45%) lesions were malignant and 145 (55%) lesions were benign. MRI was significantly more sensitive than BWBUS (p = 0.01), while BWBUS was significantly more accurate and specific than MRI (p < 0.0001). Compared with mammography, the ICDRs using BWBUS and MRI were significantly higher for oestrogen receptor-positive and triple-negative cancers, but not for human epidermal growth factor receptor 2-positive cancers. 22 additional malignant lesions in 18 patients were seen on MRI only. Surgical planning remained unchanged in 8 (44%) of those 18 patients. CONCLUSION: MRI was more sensitive than BWBUS, while BWBUS was more accurate and specific than MRI. MRI-detected additional malignant lesions did not change surgical planning in almost half of these patients. ADVANCES IN KNOWLEDGE: BWBUS may be a cost-effective and practical tool in breast cancer staging.

Breast Imaging: A Paradigm for Accountable Care Organizations.

OBJECTIVE: Accountable care organizations (ACOs) are being promoted by the Centers of Medicare Services as alternative payment models for radiology reimbursement. Because of its clinical orientation, focus on prevention, standardized reporting, quality orientation through mandatory accreditation, and value demonstration through established outcome metrics, breast imaging offers a unique paradigm for the ACO model in radiology. CONCLUSION: In radiology, breast imaging represents the paradigm for ACOs.

Dual resolution cone beam breast CT: a feasibility study.

PURPOSE: In this study, the authors investigated the feasibility of a dual resolution volume-of-interest (VOI) cone beam breast CT technique and compared two implementation approaches in terms of dose saving and scatter reduction. METHODS: With this technique, a lead VOI mask with an opening is inserted between the x-ray source and the breast to deliver x-ray exposure to the VOI while blocking x rays outside the VOI. A CCD detector is used to collect the high resolution projection data of the VOI. Low resolution cone beam CT (CBCT) images of the entire breast, acquired with a flat panel (FP) detector, were used to calculate the projection data outside the VOI with the ray-tracing reprojection method. The Feldkamp-Davis-Kress filtered backprojection algorithm was used to reconstruct the dual resolution 3D images. Breast phantoms with 180 microm and smaller microcalcifications (MCs) were imaged with both FP and FP-CCD dual resolution CBCT systems, respectively. Two approaches of implementing the dual resolution technique, breast-centered approach and VOI-centered approach, were investigated and evaluated for dose saving and scatter reduction with Monte Carlo simulation using a GEANT4 package. RESULTS: The results showed that the breast-centered approach saved more breast absorbed dose than did VOI-centered approach with similar scatter reduction. The MCs in fatty breast phantom, which were invisible with FP CBCT scan, became visible with the FP-CCD dual resolution CBCT scan. CONCLUSIONS: These results indicate potential improvement of the image quality inside the VOI with reduced breast dose both inside and outside the VOI.

Reduction in x-ray scatter and radiation dose for volume-of-interest (VOI) cone-beam breast CT--a phantom study.

With volume-of-interest (VOI) cone-beam computed tomography (CBCT) imaging, one set of projection images are acquired with the VOI collimator at a regular or high exposure level and the second set of projection images are acquired without the collimator at a reduced exposure level. The high exposure VOI scan data inside the VOI and the low exposure full-field scan data outside the VOI are then combined together to generate composite projection images for image reconstruction. To investigate and quantify scatter reduction, dose saving and image quality improvement in VOI CBCT imaging, a flat panel detector-based bench-top experimental CBCT system was built to measure the dose, the scatter-to-primary ratio (SPR), the image contrast, noise level, the contrast-to-noise ratio (CNR) and the figure of merit (FOM) in the CBCT reconstructed images for two polycarbonate cylinders simulating the small and the large phantoms. The results showed that, compared to the full field CBCT technique, radiation doses for the VOI CBCT technique were reduced by a factor of 1.20 and 1.36 for the small and the large phantoms at the phantom center, respectively, and from 2.7 to 3.0 on the edge of the phantom, respectively. Inside the VOI, the SPRs were substantially reduced by a factor of 6.6 and 10.3 for the small and the large phantoms, the contrast signals were improved by a factor of 1.35 and 1.8, and the noise levels were increased by a factor of 1.27 and 1.6, respectively. As a result, the CNRs were improved by a factor of 1.06 and 1.13 for the small and the large phantoms and the FOM improved by a factor of 1.4 and 1.7, respectively.

Feasibility of volume-of-interest (VOI) scanning technique in cone beam breast CT--a preliminary study.

This work is to demonstrate that high quality cone beam CT images can be generated for a volume of interest (VOI) and to investigate the exposure reduction effect, dose saving, and scatter reduction with the VOI scanning technique. The VOI scanning technique involves inserting a filtering mask between the x-ray source and the breast during image acquisition. The mask has an opening to allow full x-ray exposure to be delivered to a preselected VOI and a lower, filtered exposure to the region outside the VOI. To investigate the effects of increased noise due to reduced exposure outside the VOI on the reconstructed VOI image, we directly extracted the projection data inside the VOI from the full-field projection data and added additional data to the projection outside the VOI to simulate the relative noise increase due to reduced exposure. The nonuniform reference images were simulated in an identical manner to normalize the projection images and measure the x-ray attenuation factor for the object. Regular Feldkamp-Davis-Kress filtered backprojection algorithm was used to reconstruct the 3D images. The noise level inside the VOI was evaluated and compared with that of the full-field higher exposure image. Calcifications phantom and low contrast phantom were imaged. Dose reduction was investigated by estimating the dose distribution in a cylindrical water phantom using Monte Carlo simulation based Geant4 package. Scatter reduction at the detector input was also studied. Our results show that with the exposure level reduced by the VOI mask, the dose levels were significantly reduced both inside and outside the VOI without compromising the accuracy of image reconstruction, allowing for the VOI to be imaged with more clarity and helping to reduce the breast dose. The contrast-to-noise ratio inside the VOI was improved. The VOI images were not adversely affected by noisier projection data outside the VOI. Scatter intensities at the detector input were also shown to decrease significantly both inside and outside the VOI in the projection images, indicating potential improvement of image quality inside the VOI and contribution to dose reduction both inside and outside the VOI.