Automated Breast Ultrasound Research Articles

The lesion detection efficacy of deep learning on automatic breast ultrasound and factors affecting its efficacy: a pilot study.

The aim of this study was to investigate the detection efficacy of deep learning (DL) for automatic breast ultrasound (ABUS) and factors affecting its efficacy. Females who underwent ABUS and handheld ultrasound from May 2016 to June 2017 (N = 397) were enrolled and divided into training (n = 163 patients with breast cancer and 33 with benign lesions), test (n = 57) and control (n = 144) groups. A convolutional neural network was optimized to detect lesions in ABUS. The sensitivity and false positives (FPs) were evaluated and compared for different breast tissue compositions, lesion sizes, morphologies and echo patterns. In the training set, with 688 lesion regions (LRs), the network achieved sensitivities of 93.8%, 97.2% and 100%, based on volume, lesion and patient, respectively, with 1.9 FPs per volume. In the test group with 247 LRs, the sensitivities were 92.7%, 94.5% and 96.5%, respectively, with 2.4 FPs per volume. The control group, with 900 volumes, showed 0.24 FPs per volume. The sensitivity was 98% for lesions > 1 cm3, but 87% for those ≤1 cm3 (p < 0.05). Similar sensitivities and FPs were observed for different breast tissue compositions (homogeneous, 97.5%, 2.1; heterogeneous, 93.6%, 2.1), lesion morphologies (mass, 96.3%, 2.1; non-mass, 95.8%, 2.0) and echo patterns (homogeneous, 96.1%, 2.1; heterogeneous 96.8%, 2.1). DL had high detection sensitivity with a low FP but was affected by lesion size. DL is technically feasible for the automatic detection of lesions in ABUS.

Dependability of Automated Breast Ultrasound (ABUS) in Assessing Breast Imaging Reporting and Data System (BI-RADS) Category and Size of Malignant Breast Lesions Compared with Handheld Ultrasound (HHUS) and Mammography (MG).

PurposeThis study aimed to evaluate the dependability of automated breast ultrasound (ABUS) compared with handheld ultrasound (HHUS) and mammography (MG) on the Breast Imaging Reporting and Data System (BI-RADS) category and size assessment of malignant breast lesions.Patients and MethodsA total of 344 confirmed malignant lesions were recruited. All participants underwent MG, HHUS, and ABUS examinations. Agreements on the BI-RADS category were evaluated. Lesion size assessed using the three methods was compared with the size of the pathological result as the control. Regarding the four major molecular subtypes, correlation coefficients between size on imaging and pathology were also evaluated.ResultsThe agreement between ABUS and HHUS on the BI-RADS category was 86.63% (kappa = 0.77), whereas it was 32.22% (kappa = 0.10) between ABUS and MG. Imaging lesion size compared to pathologic lesion size was assessed correctly in 36.92%/52.91% (ABUS), 33.14%/48.84% (HHUS) and 33.44%/43.87% (MG), with the threshold of 3 mm/5 mm, respectively. The correlation coefficient of size of ABUS-Pathology (0.75, Spearman) was statistically higher than that of the MG-Pathology (0.58, Spearman) with P < 0.01, but not different from that of the HHUS-Pathology (0.74, Spearman) with P > 0.05. The correlation coefficient of ABUS-Pathology was statistically higher than that of MG-Pathology in the triple-negative subtype, luminal B subtype, and luminal A subtype (P<0.01).ConclusionThe agreement between ABUS and HHUS in the BI-RADS category was good, whereas that between ABUS and MG was poor. ABUS and HHUS allowed a more accurate assessment of malignant tumor size compared to MG.

Clinical Application of Automated Breast Ultrasound

Automatic breast ultrasound (ABUS) has been developed to compensate for the shortcomings of hand-held ultrasound (HHUS) and is mainly used for breast cancer screening purposes in women with dense breasts. Since 2021, ABUS has been covered by the Korean National Health Insurance System. It is important to scan the entire breast on ABUS and to identify the poor-quality images requiring re-scanning. In addition, a general understanding of the unique ABUS display mode, distinguishing benign from malignant lesions, the diagnostic performance of breast cancer screening, and the application of computer-aided detection/diagnosis systems is necessary to use ABUS efficiently. This review explores the acquisition method, image quality, and image characteristics of ABUS to improve general understanding of this procedure and its advantages over HHUS, so that ABUS can be applied efficiently in clinical practice.

Do automated breast ultrasound and tomosynthesis have an effective role in dense breast evaluation?

BackgroundMammography plays a great role in reducing breast cancer mortality as it is the standard method of breast imaging and screening. But the accuracy of mammography performance reduces in cancer detection in women with dense breast due to the summation of images and overlapping of breast tissue. ABUS and tomosynthesis both recently help to detect breast cancer in dense breasted women. This prospective study was done in the female imaging unit and approved by its research and ethical committee; all the patients did an informed consent during the period from October 2018 to March 2019. The study was conducted on 38 patients with 38 lesions subjected to digital mammography, tomosynthesis and automated breast ultrasound (ABUS), who all had dense breast in mammography.ResultsAutomated breast ultrasound (ABUS) showed 100% in all sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV) as well as accuracy, while the digital mammography tomosynthesis showed 100% in specificity, 87.5% in sensitivity, 100% in PPV, 82.4% in NPV and 92.1% accuracy.ConclusionAutomated breast ultrasound (ABUS) together with tomosynthesis makes a revolution in breast screening and detecting cancer in women with dense breasts.

Re-evaluation of high-risk breast mammography lesions by target ultrasound and ABUS of breast non-mass-like lesions

ObjectiveThe purpose of this study was to reevaluate the high-risk breast non-mass-like lesions (NMLs) in mammography (MG) by target ultrasound (US) and Automated breast ultrasonography (ABUS), and to analyze the correlation between different imaging findings and the factors influencing the classification of lesions.MethodsA total of 161 patients with 166 breast lesions were recruited in this retrospectively study. All cases were diagnosed as BI-RADS 4 or 5 by MG and as NML on ultrasound. While all NMLs underwent mammography, target US and ABUS before breast surgery or biopsy in the consistent position of breast. The imaging and pathological features of all cases were collected. All lesions were classified according to the lexion of ACR BI-RADS®.ResultsThere were significant differences between benign and malignant breast NML in all the features of target US and ABUS. US, especially ABUS, was superior to MG in determining the malignant breast NML. There was a significant difference in the detection rate of calcification between MG and Target US (P < 0.001), and there was a significant difference in the detection rate of structural distortion between ABUS and MG (P < 0.001).ConclusionsTarget US, especially ABUS, can significantly improve the sensitivity, specificity and accuracy of the diagnosis of high-risk NMLs in MG. The features of Target US and ABUS such as blood supply, hyperechogenicity, ductal changes, peripheral changes and coronal features could be employed to predict benign and malignant lesions. The coronal features of ABUS were more sensitive than those of Target HHUS in showing structural abnormalities. Target US was less effective than MG in local micro-calcification.

Outcomes of Return to Routine Screening for BI-RADS 3 Lesions Detected at Supplemental Automated Whole-Breast Ultrasound in Women With Dense Breasts: A Prospective Study.

BACKGROUND. Supplemental screening breast ultrasound (US) detects additional cancers in women with dense breasts but identifies many BI-RADS 3 lesions that result in short-term follow-up and biopsies. OBJECTIVE. The purpose of this study was to evaluate outcomes in patients recommended for return to routine screening for lesions assessed as BI-RADS 3 on supplemental automated whole-breast US. METHODS. This prospective study invited patients with BI-RADS 1 or 2 on screening mammography and breast density C or D to undergo supplemental automated breast US (ABUS). ABUS was interpreted as BI-RADS 1, 2, 3, or 0. Return to routine screening was recommended for ABUS BI-RADS 1, 2, or 3. ABUS BI-RADS 0 lesions underwent targeted handheld US. Remaining patients were followed for 2 years. Malignancy rates were compared using Fisher exact tests. RESULTS. A total of 2257 women (mean age, 58.0 ± 11.2 [SD] years) were included. Supplemental ABUS was scored as BI-RADS 1 in 1186 (52.5%) women, BI-RADS 2 in 591 (26.2%), BI-RADS 3 in 395 (17.5%), and BI-RADS 0 in 85 (3.8%). A total of 394 patients with ABUS BI-RADS 3 had 2-year follow-up, during which no cancer (0%; 95% CI, 0.0-0.9%) was diagnosed in the quadrant of the lesion. Among patients with 2-year follow-up, breast cancer was diagnosed in 4/1117 (0.4%) with ABUS BI-RADS 1, 2/556 (0.4%) with ABUS BI-RADS 2, and 2/394 (0.5%) with ABUS BI-RADS 3 (cancer in other quadrant than the lesion). Malignancy rates were not different between ABUS BI-RADS 1, 2, and 3 (p = .28). The ABUS recall rate was 3.8% (85/2257; 95% CI, 3.6-4.0%). If short-term follow-up had been recommended for ABUS BI-RADS 3, the ABUS recall rate would have been 21.3% (480/2257, 95% CI 19.6-23.0%). The biopsy rate was 0.5% (12/2257; 95% CI, 0.3-0.9%); the positive biopsy rate was 58.3% (7/12). One of seven cancers diagnosed by initial supplemental ABUS and none of eight cancers diagnosed during subsequent follow-up were node-positive cancer. CONCLUSION. Return to routine screening for ABUS BI-RADS 3 lesions results in a substantial decrease in recall rate and is unlikely to result in an adverse outcome. CLINICAL IMPACT. This prospective study supports a recommendation for routine annual follow-up for BI-RADS 3 lesions at supplemental ABUS. TRIAL REGISTRATION. ClinicalTrials.gov NCT02650778.

Cost-Benefit relationship between Hand Held Ultra-Sound (HHUS) and Automated Breast Ultra-Sound (ABUS) as a complementary scan for Mammography in detection of breast lesions

Abstract Aim of the Work to compare between HHUS and 3D ABUS, according to their benefits and limitations as a complementary scan for mammography in detection of breast lesions. Patients and Methods The study was conducted on 30 female patients with breast masses diagnosed by clinical examination or by sonomammography in Radiology Department at Al Sheikh Zayed Aal Nahian hospital and Ain Shams University Hospitals during the period from October 2018 and August 2019. Findings were confirmed by histopathological biopsy or at least 6 months follow up. Results Statistically significant difference between ABUS and HHUS time of examination in favor of ABUS with P-value &amp;gt;0.0001 and 95%CI-5.7:-4.4, as ABUS time of examination was (mean+/- SD: 4.1 +/-0.9) and HHUS was (mean +/- SD: 9.2 +/-2.0). In this study; 8 out of 30 were malignant lesions and 22 were benign findings. HHUS showed higher sensitivity than ABUS (100% versus 88%); specificity (86%, 91%), positive predictive value (73%, 78%) and negative predictive value (100%, 95%) respectively. HHUS and ABUS shows nearly perfect agreement regarding lesion classification(benign &amp;malignant).with Cohen’s kappa (κ) 0.85, and also regarding BIRADS classification With Weighted kappa 0.824. Conclusion There is a near perfect agreement between ABUS and HHUS diagnostic performance (κ) 0.85,and statistically significant difference in time of examination in favor of ABUS as it is operator independent and provide large scanned area in every single sweep and also aided with computer software, but HHUS still have higher sensitivity than ABUS and that may be attributed to ABUS limitations to evaluate axillary lymph nodes and lack of Doppler, so ABUS may serve as an effective, adjunct, screening tool to mammography and hand held Sonography.

Second-Generation 3D Automated Breast Ultrasonography (Prone ABUS) for Dense Breast Cancer Screening Integrated to Mammography: Effectiveness, Performance and Detection Rates.

In our study, we added a three-dimensional automated breast ultrasound (3D ABUS) to mammography to evaluate the performance and cancer detection rate of mammography alone or with the addition of 3D prone ABUS in women with dense breasts. Our prospective observational study was based on the screening of 1165 asymptomatic women with dense breasts who selected independent of risk factors. The results evaluated include the cancers detected between June 2017 and February 2019, and all surveys were subjected to a double reading. Mammography detected four cancers, while mammography combined with a prone Sofia system (3D ABUS) doubled the detection rate, with eight instances of cancer being found. The diagnostic yield difference was 3.4 per 1000. Mammography alone was subjected to a recall rate of 14.5 for 1000 women, while mammography combined with 3D prone ABUS resulted in a recall rate of 26.6 per 1000 women. We also observed an additional 12.1 recalls per 1000 women screened. Integrating full-field digital mammography (FFDM) with 3D prone ABUS in women with high breast density increases and improves breast cancer detection rates in a significant manner, including small and invasive cancers, and it has a tolerable impact on recall rate. Moreover, 3D prone ABUS performance results are comparable with the performance results of the supine 3D ABUS system.

Does automated breast ultrasound (ABUS) add to breast tomosynthesis (DBT) in assessment of lesions in dense breasts?

BackgroundAs mammography has its known limitations in dense breast, additional imaging is usually needed. We aimed to evaluate the role of automated breast ultrasound in addition to tomosynthesis in detection and diagnosis of breast lesions in dense breasts. Seventy patients with dense breasts subjected to full-field digital mammography (FFDM) including digital breast tomosynthesis (DBT) and automated breast ultrasound (ABUS). Both studies were evaluated by two experienced radiologists to assess breast composition, mass characterization, asymmetry, calcification, axillary lymphadenopathy, extent of disease (EOD), skin thickening, retraction, architectural distortion, and BIRADS classification. All breast masses were interpreted as above described and then correlated with final pathological diagnosis.ResultsStudy included 70 females presenting with different types of breast lesions. Eighty-two masses were detected: 53 benign (n = 53/82), 29 malignant (n = 29/82). Histopathology of the masses was reached by core biopsy (n = 30), FNAC (n = 14), and excisional biopsy (n = 11). The rest of the masses (n = 27/82) were confirmed by their characteristic sonographic appearances; 20 cases of multiple bilateral anechoic simple cysts, 7 typical fibroadenomas showed stationary course on follow-up. As regards the final BIRADS score given for both modalities, tomosynthesis showed accuracy of 93.1% in characterization of malignant masses with accuracy of 94.3% in benign masses, on the other hand automated ultrasound showed 100% accuracy in characterization of malignant masses with 98.1% accuracy in benign masses.ConclusionAdding ABUS to tomosynthesis has proven a valuable imaging tool for characterization of breast lesions in dense breasts both as screening and diagnostic tool. They proved to be more sensitive and specific than digital mammography alone in showing tissue overlap, tumor characterization, lesion margins, extent, and multiplicity of malignant lesions.

Dilated densely connected U-Net with uncertainty focus loss for 3D ABUS mass segmentation

Background and objectiveAccurate segmentation of breast mass in 3D automated breast ultrasound (ABUS) images plays an important role in qualitative and quantitative ABUS image analysis. Yet this task is challenging due to the low signal to noise ratio and serious artifacts in ABUS images, the large shape and size variation of breast masses, as well as the small training dataset compared with natural images. The purpose of this study is to address these difficulties by designing a dilated densely connected U-Net (D2U-Net) together with an uncertainty focus loss. MethodsA lightweight yet effective densely connected segmentation network is constructed to extensively explore feature representations in the small ABUS dataset. In order to deal with the high variation in shape and size of breast masses, a set of hybrid dilated convolutions is integrated into the dense blocks of the D2U-Net. We further suggest an uncertainty focus loss to put more attention on unreliable network predictions, especially the ambiguous mass boundaries caused by low signal to noise ratio and artifacts. Our segmentation algorithm is evaluated on an ABUS dataset of 170 volumes from 107 patients. Ablation analysis and comparison with existing methods are conduct to verify the effectiveness of the proposed method. ResultsExperiment results demonstrate that the proposed algorithm outperforms existing methods on 3D ABUS mass segmentation tasks, with Dice similarity coefficient, Jaccard index and 95% Hausdorff distance of 69.02%, 56.61% and 4.92 mm, respectively. ConclusionsThe proposed method is effective in segmenting breast masses on our small ABUS dataset, especially breast masses with large shape and size variations.

Pros and Cons for Automated Breast Ultrasound (ABUS): A Narrative Review.

Automated breast ultrasound (ABUS) is an ultrasound technique that tends to be increasingly used as a supplementary technique in the evaluation of patients with dense glandular breasts. Patients with dense breasts have an increased risk of developing breast cancer compared to patients with fatty breasts. Furthermore, for this group of patients, mammography has a low sensitivity in detecting breast cancers, especially if it is not associated with architectural distortion or calcifications. ABUS is a standardized examination with many advantages in both screening and diagnostic settings: it increases the detection rate of breast cancer, improves the workflow, and reduces the examination time. On the other hand, like any imaging technique, ABUS has disadvantages and even some limitations. Many disadvantages can be diminished by additional attention and training. Disadvantages regarding image acquisition are the inability to assess the axilla, the vascularization, and the elasticity of a lesion, while concerning the interpretation, the disadvantages are the artifacts due to poor positioning, lack of contact, motion or lesion related. This article reviews and discusses the indications, the advantages, and disadvantages of the method and also the sources of error in the ABUS examination.

Self-supervised learning for accelerated 3D high-resolution ultrasound imaging.

Ultrasound (US) imaging has been widely used in diagnosis, image-guided intervention, and therapy, where high-quality three-dimensional (3D) images are highly desired from sparsely acquired two-dimensional (2D) images. This study aims to develop a deep learning-based algorithm to reconstruct high-resolution (HR) 3D US images only reliant on the acquired sparsely distributed 2D images. We propose a self-supervised learning framework using cycle-consistent generative adversarial network (cycleGAN), where two independent cycleGAN models are trained with paired original US images and two sets of low-resolution (LR) US images, respectively. The two sets of LR US images are obtained through down-sampling the original US images along the two axes, respectively. In US imaging, in-plane spatial resolution is generally much higher than through-plane resolution. By learning the mapping from down-sampled in-plane LR images to original HR US images, cycleGAN can generate through-plane HR images from original sparely distributed 2D images. Finally, HR 3D US images are reconstructed by combining the generated 2D images from the two cycleGAN models. The proposed method was assessed on two different datasets. One is automatic breast ultrasound (ABUS) images from 70 breast cancer patients, the other is collected from 45 prostate cancer patients. By applying a spatial resolution enhancement factor of 3 to the breast cases, our proposed method achieved the mean absolute error (MAE) value of 0.90±0.15, the peak signal-to-noise ratio (PSNR) value of 37.88±0.88dB, and the visual information fidelity (VIF) value of 0.69±0.01, which significantly outperforms bicubic interpolation. Similar performances have been achieved using the enhancement factor of 5 in these breast cases and using the enhancement factors of 5 and 10 in the prostate cases. We have proposed and investigated a new deep learning-based algorithm for reconstructing HR 3D US images from sparely acquired 2D images. Significant improvement on through-plane resolution has been achieved by only using the acquired 2D images without any external atlas images. Its self-supervision capability could accelerate HR US imaging.

Cross-Model Attention-Guided Tumor Segmentation for 3D Automated Breast Ultrasound (ABUS) Images.

Tumor segmentation in 3D automated breast ultrasound (ABUS) plays an important role in breast disease diagnosis and surgical planning. However, automatic segmentation of tumors in 3D ABUS images is still challenging, due to the large tumor shape and size variations, and uncertain tumor locations among patients. In this paper, we develop a novel cross-model attention-guided tumor segmentation network with a hybrid loss for 3D ABUS images. Specifically, we incorporate the tumor location into a segmentation network by combining an improved 3D Mask R-CNN head into V-Net as an end-to-end architecture. Furthermore, we introduce a cross-model attention mechanism that is able to aggregate the segmentation probability map from the improved 3D Mask R-CNN to each feature extraction level in the V-Net. Then, we design a hybrid loss to balance the contribution of each part in the proposed cross-model segmentation network. We conduct extensive experiments on 170 3D ABUS from 107 patients. Experimental results show that our method outperforms other state-of-the-art methods, by achieving the Dice similarity coefficient (DSC) of 64.57%, Jaccard coefficient (JC) of 53.39%, recall (REC) of 64.43%, precision (PRE) of 74.51%, 95th Hausdorff distance (95HD) of 11.91mm, and average surface distance (ASD) of 4.63mm. Our code will be available online (https://github.com/zhouyuegithub/CMVNet).

Tumor detection using deep learning method in automated breast ultrasound

ObjectiveAccurate and efficient breast tumor detection of automated breast ultrasound (ABUS) plays a significant role for clinical diagnosis. However, reviewing thousands of ABUS slices is extremely a time-consuming task. Besides, speckle, shadowing, non-uniform contrast of certain structures and high variability of the lesion echogenicity, make tumor detection in ABUS images difficult and prone to false positive (FP) regions. To improve the tumor detection rate and reduce FP regions, a Bayesian YOLOv4 network based on Monte Carlo dropout (MC-Drop) is proposed. MethodsFirstly, some tricks in YOLOv4 network are adjusted to make it more suitable for ABUS tumor detection. Secondly, MC-Drop is applied to introduce uncertainty into the detection network and the Bayesian YOLOv4 network is constructed, which effectively reduces accidental FP regions and improves the detection rate for hard tumor region. ResultsThis method was tested on 68 volumes, including 21,624 slices (1683 tumor slices and 19,941 normal slices). It obtains a promising result with sensitivity of 0.88 and FPs/S at 0.19. ConclusionExperimental results demonstrate that our Bayesian YOLOv4 outperforms the original YOLOv4 network and some other mainstream object detectors.

Tumor classification in automated breast ultrasound (ABUS) based on a modified extracting feature network.

People can get consistent Automated Breast Ultrasound (ABUS) images due to the imaging mechanism of scanning. Therefore, it has unique advantages in breast tumor classification using artificial intelligence technology. This paper proposes a method for classifying benign and malignant breast tumors using ABUS sequence based on deep learning. First, Images of Interest (IOI) will be extracted and Region of Interest (ROI) will be cropped in ABUS sequence by two preprocessing deep learning models, Extracting-IOI model and Cropping-ROI model. Then, we propose a Shallowly Dilated Convolutional Branch Network (SDCB-Net). We combine this network with the VGG16 transfer learning network to construct a brand-new Shared Extracting Feature Network (SEF-Net) to extract ROI sequence features. Finally, the correlation features of ABUS images are extracted and integrated by using GRU Classified Network (GRUC-Net) to achieve the accurate breast tumors classification. The final results show that the accuracy of the test set for classifying benign and malignant ABUS sequence is 92.86 %. This method not only has high accuracy but also greatly improves the speed and efficiency of breast tumor classification. It has high clinical application significance that more women can discover breast tumors timely.

Automated breast US as the primary screening test for breast cancer among East Asian women aged 40–49 years: a multicenter prospective study

To prospectively evaluate the diagnostic performance of screening ABUS as the primary screening test for breast cancer among Korean women aged 40-49 years. This prospective, multicenter study included asymptomatic Korean women aged 40-49 years from three academic centers between February 2017 and October 2019. Each participant underwent ABUS without mammography, and the ABUS images were interpreted at each hospital with double-reading by two breast radiologists. Biopsy and at least 1 year of follow-up was considered the reference standard. Diagnostic performance of ABUS screening and subgroup analyses according to patient and tumor characteristics were evaluated. Reference standard data were available for 959 women. The recall rate was 9.8% (95% confidence interval [CI]: 7.9%, 11.7%; 94 of 959 women) and the cancer detection yield was 5.2 per 1000 women (95% CI: -0.6, 11.1; 5 of 959 women). There was only one interval cancer. The sensitivity was 83.3% (95% CI: 53.5%, 100%; 5 of 6 cancers) and the specificity was 90.7% (95% CI: 88.8%, 92.5%; 864 of 95. women). The positive predictive values of biopsies performed (PPV3) was 20.0% (95% CI: 4.3%, 35.7%; 5 of 25 women). Women with heterogeneous background echotexture had a higher recall rate (p = .009) and lower specificity (p = .036). Women with body mass index values < 25 kg/m2 had a higher mean recall rate (p = .046). In East Asia, screening automated breast US may be an alternative to screening mammography for detecting breast cancers in women aged 40-49 years. • Automated breast US screening for breast cancer in asymptomatic women aged 40-49 is effective with 5.2 per 1000 cancer detection yield. • Women with heterogeneous background echotexture had a higher recall rate and lower specificity. • Women with body mass index < 25 kg/m2 had a higher recall rate.

Uncertainty Aware Temporal-Ensembling Model for Semi-Supervised ABUS Mass Segmentation.

Accurate breast mass segmentation of automated breast ultrasound (ABUS) images plays a crucial role in 3D breast reconstruction which can assist radiologists in surgery planning. Although the convolutional neural network has great potential for breast mass segmentation due to the remarkable progress of deep learning, the lack of annotated data limits the performance of deep CNNs. In this article, we present an uncertainty aware temporal ensembling (UATE) model for semi-supervised ABUS mass segmentation. Specifically, a temporal ensembling segmentation (TEs) model is designed to segment breast mass using a few labeled images and a large number of unlabeled images. Considering the network output contains correct predictions and unreliable predictions, equally treating each prediction in pseudo label update and loss calculation may degrade the network performance. To alleviate this problem, the uncertainty map is estimated for each image. Then an adaptive ensembling momentum map and an uncertainty aware unsupervised loss are designed and integrated with TEs model. The effectiveness of the proposed UATE model is mainly verified on an ABUS dataset of 107 patients with 170 volumes, including 13382 2D labeled slices. The Jaccard index (JI), Dice similarity coefficient (DSC), pixel-wise accuracy (AC) and Hausdorff distance (HD) of the proposed method on testing set are 63.65%, 74.25%, 99.21% and 3.81mm respectively. Experimental results demonstrate that our semi-supervised method outperforms the fully supervised method, and get a promising result compared with existing semi-supervised methods.

Evaluation of the FUSION-X-US-II prototype to combine automated breast ultrasound and tomosynthesis

ObjectiveThe FUSION-X-US-II prototype was developed to combine 3D automated breast ultrasound (ABUS) and digital breast tomosynthesis in a single device. We evaluated the performance of ABUS and tomosynthesis in a single examination in a clinical setting.MethodsIn this prospective feasibility study, digital breast tomosynthesis and ABUS were performed using the FUSION-X-US-II prototype without any change of the breast position in patients referred for clarification of breast lesions with an indication for tomosynthesis. The tomosynthesis and ABUS images of the prototype were interpreted independently from the clinical standard by a breast diagnostics specialist. Any detected lesion was classified using BI-RADS® scores, and results of the standard clinical routine workup (gold standard) were compared to the result of the separate evaluation of the prototype images. Image quality was rated subjectively and coverage of the breast was measured.ResultsOne hundred one patients received both ABUS and tomosynthesis using the prototype. The duration of the additional ABUS acquisition was 40 to 60 s. Breast coverage by ABUS was approximately 80.0%. ABUS image quality was rated as diagnostically useful in 86 of 101 cases (85.1%). Thirty-three of 34 malignant breast lesions (97.1%) were identified using the prototype.ConclusionThe FUSION-X-US-II prototype allows a fast ABUS scan in combination with digital breast tomosynthesis in a single device integrated in the clinical workflow. Malignant breast lesions can be localized accurately with direct correlation of ABUS and tomosynthesis images. The FUSION system shows the potential to improve breast cancer screening in the future after further technical improvements.Key Points• The FUSION-X-US-II prototype allows the combination of automated breast ultrasound and digital breast tomosynthesis in a single device without decompression of the breast.• Image quality and coverage of ABUS are sufficient to accurately detect malignant breast lesions.• If tomosynthesis and ABUS should become part of breast cancer screening, the combination of both techniques in one device could offer practical and logistic advantages. To evaluate a potential benefit of a combination of ABUS and tomosynthesis in screening-like settings, further studies are needed.

The role of automated breast ultrasound in the assessment of the local extent of breast cancer.

Automated breast ultrasound (ABUS) is a non-invasive advanced ultrasound modality. The degree of extension of the cancer within the breast is very important to choose the appropriate kind of surgery/therapy. In the current work, the aim was to evaluate the role of the ABUS in the assessment of the local extent of the breast cancer before management. This is a prospective analysis that studied 562 female patients with proved breast cancers. Evaluation was in regard of the size, multiplicity, and the stromal invasion (ie, thepresence of tumor emboli or tumor masses within the stroma of the breast tissue) around the tumor. Cases were subjected to automated breast ultrasound performed in the axial and coronal planes. ABUS showed high accuracy of assessment of the tumor multiplicity (82.2%) and the stromal involvement (93.5%). There was a statistical significance (P<.001) between the ABUS and the pathology regarding the measurement of the size of the index cancer. In conclusion, ABUS could be used for determination of the intramammary extend of the breast cancer. ABUS provided accurate assessment of the peritumor stromal involvement and multiplicity of the cancer which is required to choose the proper choice of surgery.

Multi-task learning for segmentation and classification of tumors in 3D automated breast ultrasound images.

Tumor classification and segmentation are two important tasks for computer-aided diagnosis (CAD) using 3D automated breast ultrasound (ABUS) images. However, they are challenging due to the significant shape variation of breast tumors and the fuzzy nature of ultrasound images (e.g., low contrast and signal to noise ratio). Considering the correlation between tumor classification and segmentation, we argue that learning these two tasks jointly is able to improve the outcomes of both tasks. In this paper, we propose a novel multi-task learning framework for joint segmentation and classification of tumors in ABUS images. The proposed framework consists of two sub-networks: an encoder-decoder network for segmentation and a light-weight multi-scale network for classification. To account for the fuzzy boundaries of tumors in ABUS images, our framework uses an iterative training strategy to refine feature maps with the help of probability maps obtained from previous iterations. Experimental results based on a clinical dataset of 170 3D ABUS volumes collected from 107 patients indicate that the proposed multi-task framework improves tumor segmentation and classification over the single-task learning counterparts.