Addition of the presence of peritumoral edema to the known clinical-pathologic features significantly improved the association with disease recurrence (C-index, P = .042; integrated discrimination index, P < .001; continuous net reclassification improvement, P = .012). PurposeTo determine the prognostic value of peritumoral edema identified at preoperative breast magnetic resonance (MR) imaging for disease recurrence in patients with invasive breast cancer.Materials and MethodsBetween January 2011 and December 2012, 353 women (median age, 49 years; range, 27–77 years) with invasive breast cancer who had undergone preoperative MR imaging and mastectomy or breast-conserving surgery were identified. Two radiologists independently reviewed peritumoral edema on the basis of the degree of the signal intensity surrounding the tumor on T2-weighted images. The association of disease recurrence with peritumoral edema and clinical-pathologic features was assessed by using the multivariate Cox proportional hazards model and the integrated discrimination improvement (IDI) and continuous net reclassification improvement (NRI) indexes.ResultsTwenty-four patients (6.8%) had disease recurrence after 27.2 months of median follow-up. At multivariate analysis, higher N stage (hazard ratio ...

Conventional deep learning (DL) algorithm requires full supervision of annotating the region of interest (ROI) that is laborious and often biased. We aimed to develop a weakly-supervised DL algorithm that diagnosis breast cancer at ultrasound without image annotation. Weakly-supervised DL algorithms were implemented with three networks (VGG16, ResNet34, and GoogLeNet) and trained using 1000 unannotated US images (500 benign and 500 malignant masses). Two sets of 200 images (100 benign and 100 malignant masses) were used for internal and external validation sets. For comparison with fully-supervised algorithms, ROI annotation was performed manually and automatically. Diagnostic performances were calculated as the area under the receiver operating characteristic curve (AUC). Using the class activation map, we determined how accurately the weakly-supervised DL algorithms localized the breast masses. For internal validation sets, the weakly-supervised DL algorithms achieved excellent diagnostic performances, with AUC values of 0.92–0.96, which were not statistically different (allPs > 0.05) from those of fully-supervised DL algorithms with either manual or automated ROI annotation (AUC, 0.92–0.96). For external validation sets, the weakly-supervised DL algorithms achieved AUC values of 0.86–0.90, which were not statistically different (Ps > 0.05) or higher (P = 0.04, VGG16 with automated ROI annotation) from those of fully-supervised DL algorithms (AUC, 0.84–0.92). In internal and external validation sets, weakly-supervised algorithms could localize 100% of malignant masses, except for ResNet34 (98%). The weakly-supervised DL algorithms developed in the present study were feasible for US diagnosis of breast cancer with well-performing localization and differential diagnosis.

Peritumoral edema (PE) of breast cancer at T2-weighted MR images is considered a poor prognostic sign and may represent the microenvironment surrounding the tumor; however, its histopathological mechanism remains unclear. The purpose of the study was to identify and describe detailed histopathological characteristics associated with PE at preoperative breast MRI in breast cancer patients. This retrospective study included breast cancer patients who had undergone preoperative MRI and surgery between January 2011 and December 2012. Two radiologists determined the presence of PE in consensus based on the signal intensity surrounding the tumor at T2-weighted images. The following detailed histopathological characteristics were reviewed by two breast pathologists using four-tiered grades; lymphovascular invasion, vessel ectasia, stromal fibrosis, growth pattern, and tumor budding. Tumor necrosis and tumor infiltrating lymphocytes were assessed using a percent scale. Baseline clinicopathological characteristics, including age and histologic grade, were collected. The associations between detailed histopathologic characteristics and PE were examined using multivariable logistic regression with odds ratio (OR) calculation. A total of 136 women (median age, 49 ± 9 years) were assessed; among them 34 (25.0%) had PE. After adjustment of baseline clinicopathological characteristics that were significantly associated with PE (age, T stage, N stage, histologic grade, and subtype, all Ps < 0.05), lymphovascular invasion (P = 0.009), vessel ectasia (P = 0.021), stromal fibrosis (P = 0.024), growth pattern (P = 0.036), and tumor necrosis (P < 0.001) were also associated with PE. In comparison with patients without PE, patients with PE were more likely to have a higher degree of lymphovascular invasion (OR, 2.9), vessel ectasia (OR, 3.3), stromal fibrosis (OR, 2.5), lesser degree of infiltrative growth pattern (OR, 0.4), and higher portion of tumor necrosis (OR, 1.4). PE of breast cancer at MRI is associated with detailed histopathological characteristics of lymphovascular invasion, vessel ectasia, stromal fibrosis, growth pattern, and tumor necrosis, suggesting a relevance for tumor microenvironment.

Although breast ultrasonography is the mainstay modality for differentiating between benign and malignant breast masses, it has intrinsic problems with false positives and substantial interobserver variability. Artificial intelligence (AI), particularly with deep learning models, is expected to improve workflow efficiency and serve as a second opinion. AI is highly useful for performing three main clinical tasks in breast ultrasonography: detection (localization/ segmentation), differential diagnosis (classification), and prognostication (prediction). This article provides a current overview of AI applications in breast ultrasonography, with a discussion of methodological considerations in the development of AI models and an up-to-date literature review of potential clinical applications.

Recently, the amount of attention paid towards convolutional neural networks (CNN) in medical image analysis has rapidly increased since they can analyze and classify images faster and more accurately than human abilities. As a result, CNNs are becoming more popular and play a role as a supplementary assistant for healthcare professionals. Using the CNN on portable medical devices can enable a handy and accurate disease diagnosis. Unfortunately, however, the CNNs require high-performance computing resources as they involve a significant amount of computation to process big data. Thus, they are limited to being used on portable medical devices with limited computing resources. This paper discusses the network quantization techniques that reduce the size of CNN models and enable fast CNN inference with an energy-efficient CNN accelerator integrated into recent mobile processors. With extensive experiments, we show that the quantization technique reduces inference time by 97% on the mobile system integrating a CNN acceleration engine.