Faculty Bibliography

Recent advances in deep learning for object recognition in natural images has prompted a surge of interest in applying a similar set of techniques to medical images. Most of the initial attempts largely focused on replacing the input to such a deep convolutional neural network from a natural image to a medical image. This, however, does not take into consideration the fundamental differences between these two types of data. More specifically, detection or recognition of an anomaly in medical images depends significantly on fine details, unlike object recognition in natural images where coarser, more global structures matter more. This difference makes it inadequate to use the existing deep convolutional neural networks architectures, which were developed for natural images, because they rely on heavily downsampling an image to a much lower resolution to reduce the memory requirements. This hides details necessary to make accurate predictions for medical images. Furthermore, a single exam in medical imaging often comes with a set of different views which must be seamlessly fused in order to reach a correct conclusion. In our work, we propose to use a multi-view deep convolutional neural network that handles a set of more than one high-resolution medical image. We evaluate this network on large-scale mammography-based breast cancer screening (BI-RADS prediction) using 103 thousand images. We focus on investigating the impact of training set sizes and image sizes on the prediction accuracy. Our results highlight that performance clearly increases with the size of training set, and that the best performance can only be achieved using the images in the original resolution. This suggests the future direction of medical imaging research using deep neural networks is to utilize as much data as possible with the least amount of potentially harmful preprocessing

OBJECTIVES: This study was performed to determine the frequency, predictors, and outcomes of ultrasound (US) correlates for non-mass enhancement. METHODS: From January 2005 to December 2011, a retrospective review of 5837 consecutive breast magnetic resonance imaging examinations at our institution identified 918 non-mass enhancing lesions for which follow-up or biopsy was recommended. Retrospective review of the images identified 879 of 918 lesions (96%) meeting criteria for non-mass enhancement. Patient demographics, pathologic results, and the presence of an adjacent landmark were recorded. Targeted US examinations were recommended for 331 of 879 cases (38%), and 284 of 331 women (86%) underwent US evaluations. RESULTS: The US correlate rate for non-mass enhancement was 23% (64 of 284). An adjacent landmark was significantly associated with a US correlate (P < .001). Biopsy was recommended for 43 of 64 correlates (67%). Ultrasound-guided biopsy was performed on 39 of 43 (91%); 7 of 39 (18%) were malignant. No correlate was seen for 220 of 284 lesions (77%). At magnetic resonance imaging-guided biopsy, 14 of 117 (12%) were malignancies. For all biopsied non-mass enhancements, the malignancy rate was 18% (55 of 308) and was significantly more prevalent in the setting of a known index cancer (P < .001), older age (P < .001), the presence of a landmark (P = .002), and larger lesion size (P = .019). CONCLUSIONS: Non-mass enhancement with an adjacent landmark is more likely to have a US correlate compared to non-mass enhancement without an adjacent landmark. Non-mass enhancement in the setting of a known index cancer, older age, a landmark, and larger lesion size is more likely to be malignant. However, no statistical difference was detected in the rate of malignancy between non-mass enhancement with (18%) or without (12%) a correlate. Absence of a correlate does not obviate the need to biopsy suspicious non-mass enhancement.

PURPOSE: To compare background parenchymal enhancement (BPE) over time in patients with and without breast cancer. MATERIALS AND METHODS: This retrospective Institutional Review Board (IRB)-approved, Health Insurance Portability and Accountability Act (HIPAA)-compliant study included 116 women (25-84 years, mean 54 years) with breast cancer who underwent breast magnetic resonance imaging at 3T between 1/2/2009 and 12/29/2009 and 116 age and date-of-exam-matched women without breast cancer (23-84 years, mean 51 years). Two independent, blinded readers (R1, R2) recorded BPE (minimal, mild, moderate, marked) at three times (100, 210, and 320 seconds postcontrast). Subsequent cancers were diagnosed in 9/96 control patients with follow up (12.6-93.0 months, mean 63.6 months). Exact Mann-Whitney, Fisher's exact, and McNemar tests were performed. RESULTS: Mean BPE was not found to be different between patients with and without breast cancer at any time (P = 0.36-0.64). At time 2 as compared with time 1, there were significantly more patients, both with and without breast cancer, with BPE >minimal (R1: 90 vs. 41 [P < 0.001] and 81 vs. 36 [P < 0.001]; R2: 84 vs. 52 [P < 0.001] and 79 vs. 43 [P < 0.001]) and BPE >mild (R1: 59 vs. 10 [P < 0.001] and 47 vs. 13 [P < 0.001]; R2: 49 vs. 12 [P < 0.001] and 41 vs. 18 [P < 0.001]). BPE changes between times 2 and 3 were not significant (P = 0.083-1.0). Odds ratios for control patients developing breast cancer were significant only for R2 and ranged up to 7.67 (1.49, 39.5; P < 0.01) for BPE >mild at time 2. CONCLUSION: BPE changes between the first and second postcontrast scans and stabilizes thereafter in most patients. Further investigation into the most clinically relevant timepoint for BPE assessment is warranted. J. Magn. Reson. Imaging 2016.

PURPOSE: To explore the application of diffusion tensor imaging (DTI) for breast tissue and breast pathologies using a stimulated-echo acquisition mode (STEAM) with variable diffusion times. MATERIALS AND METHODS: In this Health Insurance Portability and Accountability Act-compliant study, approved by the local institutional review board, eight patients and six healthy volunteers underwent an MRI examination at 3 Tesla including STEAM-DTI with several diffusion times ranging from 68.5 to 902.5 ms. A DTI model was fitted to the data for each diffusion time, and parametric maps of mean diffusivity, fractional anisotropy, axial diffusivity, and radial diffusivity were computed for healthy fibroglandular tissue (FGT) and lesions. The median value of radial diffusivity for FGT was fitted to a linear decay to obtain an estimation of the surface-to-volume ratio, from which the radial diameter was calculated. RESULTS: For healthy FGT, radial diffusivity presented a linear decay with the square root of the diffusion time resulting in a range of estimated radial diameters from 202 to 496 microm, while axial diffusivity presented a nearly time-independent diffusion. Residual fat signal was reduced at longer diffusion times due to the shorter T1 of fat. Residual fat signal to the overall signal in the healthy volunteers' FGT was found to range from 2.39% to 2.55% (shortest mixing time), and from 0.40% to 0.51% (longest mixing time) for the b500 images. CONCLUSION: The use of variable diffusion times may provide an in vivo noninvasive tool to probe diffusion lengths in breast tissue and breast pathology, and might aid by improving fat suppression at longer diffusion times. J. Magn. Reson. Imaging 2016.

OBJECTIVE: The purpose of this study was to determine the rate, characteristics, and outcomes of discordant MRI-guided vacuum-assisted biopsy (VAB) in women with suspected breast cancer. MATERIALS AND METHODS: This retrospective study reviewed 1314 MRI-guided VABs performed in 1211 women between 2007 and 2013 and yielded 25 discordant results in 24 women. MRI characteristics; BI-RADS assessments; whether the lesion was missed, partially sampled, or excised at biopsy; and biopsy and surgical pathology results were reviewed. Statistical analyses were performed using Fisher exact and Mann-Whitney U tests. RESULTS: Among 1314 lesions that underwent MRI-guided VAB, 25 results were discordant (1.9%; 95% CI, 1.2-2.8%), and nine lesions with discordant results (36.0%, 95% CI, 18.5-56.9%) were malignant at surgical excision (three invasive ductal carcinoma and six ductal carcinoma in situ). There was no significant association between malignancy and lesion type, size, enhancement pattern, BI-RADS assessment, or clinical indication. Forty-four percent (11/25) of discordant lesions were missed, 48.0% (12/25) were partially sampled, and 8.0% (2/25) appeared to have been excised. Of the nine malignant lesions, 44.4% (4/9) discordant malignant lesions were missed, 44.4% (4/9) were partially sampled, and 11.1% (1/9) appeared to have been excised. Lesion sizes and types were similar in the missed and partially excised groups. CONCLUSION: The potential for false-negative results at MRI-guided VAB underscores the importance of radiologic-histologic correlation and imaging review after biopsy. Rebiopsy or excision in discordant cases is therefore recommended.

Breast magnetic resonance (MR) imaging is the only breast imaging modality that consistently encompasses extramammary structures in the thorax and upper abdomen. Incidental extramammary findings on breast MR images of patients with a history of breast cancer or other malignancies are significantly more likely to be malignant and may affect staging and treatment. An understanding of the frequency, distribution, and context of extramammary findings on breast MR images and a familiarity with common and uncommon sites of breast cancer metastasis inform the differential diagnosis and prompt the appropriate diagnostic next step, to differentiate benign from malignant findings. High-yield organ systems on breast MR images, as reflected by a high positive predictive value for malignancy, are correlated with known distant sites of breast cancer metastasis in the bone, lung, liver, and lymph nodes. Staging is considered when disease involves the skin and chest wall. Unusual sites of breast cancer metastasis from invasive lobular carcinoma are discussed, including the gastrointestinal tract, peritoneum, and adrenal glands. Nonmalignant clinically important findings involving the cardiovascular and gastrointestinal systems are reviewed, and potential pitfalls in diagnosis and interpretation are highlighted. A consistently systematic diagnostic approach is emphasized for identifying extramammary abnormalities on breast MR images. All things considered, the radiologist should be able to improve diagnostic sensitivity and specificity while interpreting extramammary findings on breast MR images. (c)RSNA, 2017.

OBJECTIVE: To examine the prognostic capabilities of intravoxel incoherent motion (IVIM) metrics and their ability to predict response to neoadjuvant treatment (NAT). Additionally, to observe changes in IVIM metrics between pre- and post-treatment MRI. METHODS: This IRB-approved, HIPAA-compliant retrospective study observed 31 breast cancer patients (32 lesions). Patients underwent standard bilateral breast MRI along with diffusion-weighted imaging before and after NAT. Six patients underwent an additional IVIM-MRI scan 12-14 weeks after initial scan and 2 cycles of treatment. In addition to apparent diffusion coefficients (ADC) from monoexponential decay, IVIM mean values (tissue diffusivity Dt, perfusion fraction fp, and pseudodiffusivity Dp) and histogram metrics were derived using a biexponential model. An additional filter identified voxels of highly vascular tumor tissue (VTT), excluding necrotic or normal tissue. Clinical data include histology of biopsy and clinical response to treatment through RECIST assessment. Comparisons of treatment response were made using Wilcoxon rank-sum tests. RESULTS: Average, kurtosis, and skewness of pseudodiffusion Dp significantly differentiated RECIST responders from nonresponders. ADC and Dt values generally increased ( approximately 70%) and VTT% values generally decreased ( approximately 20%) post-treatment. CONCLUSION: Dp metrics showed prognostic capabilities; slow and heterogeneous pseudodiffusion offer poor prognosis. Baseline ADC/Dt parameters were not significant predictors of response. This work suggests that IVIM mean values and heterogeneity metrics may have prognostic value in the setting of breast cancer NAT.