Classification Of Breast Lesions Research Articles

Explainable DCNN Decision Framework for Breast Lesion Classification from Ultrasound Images Based on Cancer Characteristics

In recent years, deep convolutional neural networks (DCNNs) have shown promising performance in medical image analysis, including breast lesion classification in 2D ultrasound (US) images. Despite the outstanding performance of DCNN solutions, explaining their decisions remains an open investigation. Yet, the explainability of DCNN models has become essential for healthcare systems to accept and trust the models. This paper presents a novel framework for explaining DCNN classification decisions of lesions in ultrasound images using the saliency maps linking the DCNN decisions to known cancer characteristics in the medical domain. The proposed framework consists of three main phases. First, DCNN models for classification in ultrasound images are built. Next, selected methods for visualization are applied to obtain saliency maps on the input images of the DCNN models. In the final phase, the visualization outputs and domain-known cancer characteristics are mapped. The paper then demonstrates the use of the framework for breast lesion classification from ultrasound images. We first follow the transfer learning approach and build two DCNN models. We then analyze the visualization outputs of the trained DCNN models using the EGrad-CAM and Ablation-CAM methods. We map the DCNN model decisions of benign and malignant lesions through the visualization outputs to the characteristics such as echogenicity, calcification, shape, and margin. A retrospective dataset of 1298 US images collected from different hospitals is used to evaluate the effectiveness of the framework. The test results show that these characteristics contribute differently to the benign and malignant lesions’ decisions. Our study provides the foundation for other researchers to explain the DCNN classification decisions of other cancer types.

Abstract PO1-07-03: An Accurate Breast Cancer Diagnosis Method Using a Multi-Feature Fusion Neural Network on Contrast-Enhanced Spectral Mammography

Abstract Objectives: Our study aims to develop an accurate and comprehensive deep neural network capable of classifying CESM images to aid in the early detection and diagnosis of breast cancer in clinical settings. Methods: We enrolled diagnostic CESM examinations conducted between January 1, 2019 and January 17, 2021. We developed and tested the performance of a multi-feature fusion network for breast lesion classification by combining low-energy (LE) and dual-energy subtracted (DES) images, dual-view, and bilateral information using a large and diverse CESM dataset. We have evaluated the ability of the proposed network to generalize on external datasets and the results were reported using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. Results: In the study period 1973CESMs were included. Mean age was 53 years ± 12 (standard deviation). In the internal test dataset, the model with CESM (LE +DES) inputs combined not only bilateral information (left and right breasts) but also dualview information (CC and MLO), achieved best diagnosis performance with AUC of 0.96 [95% confidence interval (CI): 0.95–0.97], accuracy of 0.9 [95% confidence interval (CI): 0.88–0.92], sensitivity of 0.84 [95% confidence interval (CI): 0.80–0.88] and specificity of 0.93 [95% confidence interval (CI): 0.91–0.95], while the model with LE inputs only got an AUC of 0.94 [95% (CI): 0.93–0.95], accuracy of 0.88 [95% confidence interval (CI): 0.86–0.90], sensitivity of 0.79 [95% confidence interval (CI): 0.74–0.83] and specificity of 0.93 [95% confidence interval (CI): 0.91–0.95] (external dataset: an AUC of 0.90 [95% (CI): 0.85–0.94], accuracy of 0.84 [95% confidence interval (CI): 0.78–0.89], sensitivity of 0.77 [95% confidence interval (CI): 0.61–0.88] and specificity of 0.87 [95% confidence interval (CI):0.80–0.92]). Conclusion: CESM is a promising technique in breast cancer diagnosis due to its high feasibility and potential. The results demonstrate that left-right breast fusion and dual-view fusion are helpful for accurate diagnosis on CESM images. Specifically, dual-energy subtracted (DES) images generated by CESM improve diagnostic accuracy when compared to low-energy (LE) images alone. Keywords: Contrast-enhanced spectral mammography; Deep neural network; Breast cancer diagnosis; Multi-feature fusion Citation Format: Yipeng Song, Wei Jiang. An Accurate Breast Cancer Diagnosis Method Using a Multi-Feature Fusion Neural Network on Contrast-Enhanced Spectral Mammography [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-07-03.

A novel hybrid deep CNN model for breast cancer classification using Lipschitz-based image augmentation and recursive feature elimination

Breast cancer is the leading cause of death for women worldwide. Despite its difficulties, the histopathological diagnosis must be used because modern imaging techniques cannot differentiate between benign and malignant tumors. This article introduces a hybrid deep learning method with recursive feature elimination for categorizing histopathology images of breast lesions as benign or malignant, using standard datasets, to increase diagnostic accuracy. The proposed four-stage model includes several novel techniques. First, a modified Lipschitz-based image augmentation technique is utilized to ensure a proportionate representation of various sample classes. K-fold cross-validation is employed to mitigate the bias that emerges from the selection of training and testing datasets. Then, for efficient feature extraction, a hybrid convolutional neural network (CNN) approach is used by integrating three CNNs (PResNet-34, FE-VGG-16, and M-AlexNet). With the proposed universum twin support vector machine-based recursive feature elimination technique, the initial 1,000 features extracted by each deep model are further reduced to 100. Finally, the deep features are subsequently combined into a set of 300 features, which are then fed into machine learning classifiers such as decision tree, k-nearest neighborhood, linear discriminant analysis, linear regression, and support vector machine. The results demonstrate that the proposed hybrid CNN model outperforms state-of-the-art techniques in accurately classifying breast lesions as benign or malignant, with an accuracy of 99.99 %, sensitivity of 99.56 %, specificity of 99.53 %, precision of 99.55 %, and F1 score of 99.77 %.

Harnessing the power of radiomics and deep learning for improved breast cancer diagnosis with multiparametric breast mammography

Breast cancer, with its high mortality, faces diagnostic challenges due to variability in mammography quality and breast densities, leading to inconsistencies in radiological interpretations. Computer-aided diagnostic (CAD) systems, while helpful, struggle with accurately interpreting lesion characteristics such as morphology, density, and size. To address this, our study developed advanced deep-learning algorithms to improve the detection, localization, risk assessment, and classification of breast lesions, aiming to reduce false positives on human intervention and tackle slow convergence rates. Key innovations of the approach include preprocessing techniques, advanced filtering, and data augmentation strategies to optimize model performance, mitigating over- and under-fitting concerns. A significant development is the Chaotic Leader Selective Filler Swarm Optimization (cLSFSO) method, which effectively detects breast-dense lesions by extracting textural and statistical features. Additionally, the study adapted deep learning models like modified VGGNet and SE-ResNet152 through transfer learning, significantly enhancing their capability to distinguish between normal and suspicious mammography regions. The study also introduces hybrid deep neural network-based approaches, including CNN+LSTM and CNN+SVM, for diagnosing and grading cancerous polyps from the pre-segmented ROIs. Besides, the transfer learning paradigm is employed to boost the efficacy in classifying breast masses and reducing computing time by modifying the final layer of the proposed pre-trained models. The integration of Grad-CAM techniques further refines our analysis, leading to more accurate assessments and improved diagnosis of breast anomalies. Evaluated using benchmark and private datasets, our algorithms demonstrated a sensitivity of 0.99 and an overall AUC of 0.99, indicating significant improvements in mammogram analysis. These advancements aid radiologists, potentially improving patient outcomes and contributing to medical imaging and AI in healthcare.

Real-time breast lesion classification combining diffuse optical tomography frequency domain data and BI-RADS assessment.

Ultrasound (US)-guided diffuse optical tomography (DOT) has demonstrated potential for breast cancer diagnosis, in which real-time or near real-time diagnosis with high accuracy is desired. However, DOT's relatively slow data processing and image reconstruction speeds have hindered real-time diagnosis. Here, we propose a real-time classification scheme that combines US breast imaging reporting and data system (BI-RADS) readings and DOT frequency domain measurements. A convolutional neural network is trained to generate malignancy probability scores from DOT measurements. Subsequently, these scores are integrated with BI-RADS assessments using a support vector machine classifier, which then provides the final diagnostic output. An area under the receiver operating characteristic curve of 0.978 is achieved in distinguishing between benign and malignant breast lesions in patient data without image reconstruction.

CAM-QUS guided self-tuning modular CNNs with multi-loss functions for fully automated breast lesion classification in ultrasound images

Objective. Breast cancer is the major cause of cancer death among women worldwide. Deep learning-based computer-aided diagnosis (CAD) systems for classifying lesions in breast ultrasound images can help materialise the early detection of breast cancer and enhance survival chances. Approach. This paper presents a completely automated BUS diagnosis system with modular convolutional neural networks tuned with novel loss functions. The proposed network comprises a dynamic channel input enhancement network, an attention-guided InceptionV3-based feature extraction network, a classification network, and a parallel feature transformation network to map deep features into quantitative ultrasound (QUS) feature space. These networks function together to improve classification accuracy by increasing the separation of benign and malignant class-specific features and enriching them simultaneously. Unlike the categorical crossentropy (CCE) loss-based traditional approaches, our method uses two additional novel losses: class activation mapping (CAM)-based and QUS feature-based losses, to capacitate the overall network learn the extraction of clinically valued lesion shape and texture-related properties focusing primarily the lesion area for explainable AI (XAI). Main results. Experiments on four public, one private, and a combined breast ultrasound dataset are used to validate our strategy. The suggested technique obtains an accuracy of 97.28%, sensitivity of 93.87%, F1-score of 95.42% on dataset 1 (BUSI), and an accuracy of 91.50%, sensitivity of 89.38%, and F1-score of 89.31% on the combined dataset, consisting of 1494 images collected from hospitals in five demographic locations using four ultrasound systems of different manufacturers. These results outperform techniques reported in the literature by a considerable margin. Significance. The proposed CAD system provides diagnosis from the auto-focused lesion area of B-mode BUS images, avoiding the explicit requirement of any segmentation or region of interest extraction, and thus can be a handy tool for making accurate and reliable diagnoses even in unspecialized healthcare centers.

Spatiotemporal analysis of contrast-enhanced ultrasound for differentiating between malignant and benign breast lesions.

The aim of this study was to apply spatiotemporal analysis of contrast-enhanced ultrasound (CEUS) loops to quantify the enhancement heterogeneity for improving the differentiation between benign and malignant breast lesions. This retrospective study included 120 women (age range, 18-82 years; mean, 52 years) scheduled for ultrasound-guided biopsy. With the aid of brightness-mode images, the border of each breast lesion was delineated in the CEUS images. Based on visual evaluation and quantitative metrics, the breast lesions were categorized into four grades of different levels of contrast enhancement. Grade-1 (hyper-enhanced) and grade-2 (partly-enhanced) breast lesions were included in the analysis. Four parameters reflecting enhancement heterogeneity were estimated by spatiotemporal analysis of neighboring time-intensity curves (TICs). By setting the threshold on mean parameter, the diagnostic performance of the four parameters for differentiating benign and malignant lesions was evaluated. Sixty-four of the 120 patients were categorized as grade 1 or 2 and used for estimating the four parameters. At the pixel level, mutual information and conditional entropy present significantly different values between the benign and malignant lesions (p < 0.001 in patients of grade 1, p = 0.002 in patients of grade 1 or 2). For the classification of breast lesions, mutual information produces the best diagnostic performance (AUC = 0.893 in patients of grade 1, AUC = 0.848 in patients of grade 1 or 2). The proposed spatiotemporal analysis for assessing the enhancement heterogeneity shows promising results to aid in the diagnosis of breast cancer by CEUS. The proposed spatiotemporal method can be developed as a standardized software to automatically quantify the enhancement heterogeneity of breast cancer on CEUS, possibly leading to the improved diagnostic accuracy of differentiation between benign and malignant lesions. • Advanced spatiotemporal analysis of ultrasound contrast-enhanced loops for aiding the differentiation of malignant or benign breast lesions. • Four parameters reflecting the enhancement heterogeneity were estimated in the hyper- and partly-enhanced breast lesions by analyzing the neighboring pixel-level time-intensity curves. • For the classification of hyper-enhanced breast lesions, mutual information produces the best diagnostic performance (AUC = 0.893).

Using the Kaiser Score as a clinical decision rule for breast lesion classification: Does computer-assisted curve type analysis improve diagnosis?

PurposeWe aimed to investigate the effect of using visual or automatic enhancement curve type assessment on the diagnostic performance of the Kaiser Score (KS), a clinical decision rule for breast MRI. MethodThis IRB-approved retrospective study analyzed consecutive conventional BI-RADS 0, 4 or 5 patients who underwent biopsy after 1.5T breast MRI according to EUSOBI recommendations between 2013 and 2015. The KS includes five criteria (spiculations; signal intensity (SI)-time curve type; margins of the lesion; internal enhancement; and presence of edema) resulting in scores from 1 (=lowest) to 11 (=highest risk of breast cancer). Enhancement curve types (Persistent, Plateau or Wash-out) were assessed by two radiologists independently visually and using a pixel-wise color-coded computed parametric map of curve types. KS diagnostic performance differences between readings were compared by ROC analysis. ResultsIn total 220 lesions (147 benign, 73 malignant) including mass (n = 148) and non-mass lesions (n = 72) were analyzed. KS reading performance in distinguishing benign from malignant lesions did not differ between visual analysis and parametric map (P = 0.119; visual: AUC 0.875, sensitivity 95 %, specificity 63 %; and map: AUC 0.901, sensitivity 97 %, specificity 65 %). Additionally, analyzing mass and non-mass lesions separately, showed no difference between parametric map based and visual curve type-based KS analysis as well (P = 0.130 and P = 0.787). ConclusionsThe performance of the Kaiser Score is largely independent of the curve type assessment methodology, confirming its robustness as a clinical decision rule for breast MRI in any type of breast lesion in clinical routine.

Joint localization and classification of breast masses on ultrasound images using an auxiliary attention-based framework.

Multi-task learning (MTL) methods have been extensively employed for joint localization and classification of breast lesions on ultrasound images to assist in cancer diagnosis and personalized treatment. One typical paradigm in MTL is a shared trunk network architecture. However, such a model design may suffer information-sharing conflicts and only achieve suboptimal performance for individual tasks. Additionally, the model relies on fully-supervised learning methodologies, imposing heavy burdens on data annotation. In this study, we propose a novel joint localization and classification model based on attention mechanisms and a sequential semi-supervised learning strategy to address these challenges. Our proposed framework offers three primary advantages. First, a lesion-aware network with multiple attention modules is designed to improve model performance on lesion localization. An attention-based classifier explicitly establishes correlations between the two tasks, alleviating information-sharing conflicts while leveraging location information to assist in classification. Second, a two-stage sequential semi-supervised learning strategy is designed for model training to achieve optimal performance on both tasks and substantially reduces the need for data annotation. Third, the asymmetric and modular model architecture allows for the flexible interchangeability of individual components, rendering the model adaptable to various applications. Experimental results from two different breast ultrasound image datasets under varied conditions have demonstrated the effectiveness of the proposed method. Furthermore, we conduct comprehensive investigations into the impacts of various factors on model performance, gaining in-depth insights into the mechanism of our proposed framework. The code is available at https://github.com/comp-imaging-sci/lanet-bus.git.

Clinical Breast MRI-based Radiomics for Distinguishing Benign and Malignant Lesions: An Analysis of Sequences and Enhanced Phases.

Previous studies have used different imaging sequences and different enhanced phases for breast lesion calsification in radiomics. The optimal sequence and contrast enhanced phase is unclear. To identify the optimal magnetic resonance imaging (MRI) radiomics model for lesion clarification, and to simulate its incremental value for multiparametric MRI (mpMRI)-guided biopsy. Retrospective. 329 female patients (138 malignant, 191 benign), divided into a training set (first site, n = 192) and an independent test set (second site, n = 137). 3.0-T, fast spoiled gradient-echo and fast spin-echo T1-weighted imaging (T1WI), fast spin-echo T2-weighted imaging (T2WI), echo-planar diffusion-weighted imaging (DWI), and fast spoiled gradient-echo contrast-enhanced MRI (CE-MRI). Two breast radiologists with 3 and 10 years' experience developed radiomics model on CE-MRI, CE-MRI + DWI, CE-MRI + DWI + T2WI, CE-MRI + DWI + T2WI + T1WI at each individual phase (P) and for multiple combinations of phases. The optimal radiomics model (Rad-score) was identified as having the highest area under the receiver operating characteristic curve (AUC) in the test set. Specificity was compared between a traditional mpMRI model and an integrated model (mpMRI + Rad-score) at sensitivity >98%. Wilcoxon paired-samples signed rank test, Delong test, McNemar test. Significance level was 0.05 and Bonferroni method was used for multiple comparisons (P = 0.007, 0.05/7). For radiomics models, CE-MRI/P3 + DWI + T2WI achieved the highest performance in the test set (AUC = 0.888, 95% confidence interval: 0.833-0.944). The integrated model had significantly higher specificity (55.3%) than the mpMRI model (31.6%) in the test set with a sensitivity of 98.4%. The CE-MRI/P3 + DWI + T2WI model is the optimized choice for breast lesion classification in radiomics, and has potential to reduce benign biopsies (100%-specificity) from 68.4% to 44.7% while retaining sensitivity >98%. 3 TECHNICAL EFFICACY: Stage 2.

Breast lesion classification from mammograms using deep neural network and test-time augmentation

Breast lesion classification from mammograms using deep neural network and test-time augmentation

ENAS-B: Combining ENAS With Bayesian Optimization for Automatic Design of Optimal CNN Architectures for Breast Lesion Classification From Ultrasound Images.

Efficient Neural Architecture Search (ENAS) is a recent development in searching for optimal cell structures for Convolutional Neural Network (CNN) design. It has been successfully used in various applications including ultrasound image classification for breast lesions. However, the existing ENAS approach only optimizes cell structures rather than the whole CNN architecture nor its trainable hyperparameters. This paper presents a novel framework for automatic design of CNN architectures by combining strengths of ENAS and Bayesian Optimization in two-folds. Firstly, we use ENAS to search for optimal normal and reduction cells. Secondly, with the optimal cells and a suitable hyperparameter search space, we adopt Bayesian Optimization to find the optimal depth of the network and optimal configuration of the trainable hyperparameters. To test the validity of the proposed framework, a dataset of 1522 breast lesion ultrasound images is used for the searching and modeling. We then evaluate the robustness of the proposed approach by testing the optimized CNN model on three external datasets consisting of 727 benign and 506 malignant lesion images. We further compare the CNN model with the default ENAS-based CNN model, and then with CNN models based on the state-of-the-art architectures. The results (error rate of no more than 20.6% on internal tests and 17.3% on average of external tests) show that the proposed framework generates robust and light CNN models.

BUS-BRA: A breast ultrasound dataset for assessing computer-aided diagnosis systems.

Computer-aided diagnosis (CAD) systems on breast ultrasound (BUS) aim to increase the efficiency and effectiveness of breast screening, helping specialists to detect and classify breast lesions. CAD system development requires a set of annotated images, including lesion segmentation, biopsy results to specify benign and malignant cases, and BI-RADS categories to indicate the likelihood of malignancy. Besides, standardized partitions of training, validation, and test sets promote reproducibility and fair comparisons between different approaches. Thus, we present a publicly available BUS dataset whose novelty is the substantial increment of cases with the above-mentioned annotations and the inclusion of standardized partitions to objectively assess and compare CADsystems. The BUS dataset comprises1875 anonymized images from1064 female patients acquired via four ultrasound scanners during systematic studies at the National Institute of Cancer (Rio de Janeiro, Brazil). The dataset includes biopsy-proven tumors divided into722 benign and342 malignant cases. Besides, a senior ultrasonographer performed a BI-RADS assessment in categories2 to5. Additionally, the ultrasonographer manually outlined the breast lesions to obtain ground truth segmentations. Furthermore, 5- and 10-fold cross-validation partitions are provided to standardize the training and test sets to evaluate and reproduce CAD systems. Finally, to validate the utility of the BUS dataset, an evaluation framework is implemented to assess the performance of deep neural networks for segmenting and classifying breastlesions. The BUS dataset is publicly available for academic and research purposes through an open-access repository under the name BUS-BRA: A Breast Ultrasound Dataset for Assessing CAD Systems. BUS images and reference segmentations are saved in Portable Network Graphic (PNG) format files, and the dataset information is stored in separate Comma-Separated Value (CSV)files. The BUS-BRA dataset can be used to develop and assess artificial intelligence-based lesion detection and segmentation methods, and the classification of BUS images into pathological classes and BI-RADS categories. Other potential applications include developing image processing methods like despeckle filtering and contrast enhancement methods to improve image quality and feature engineering for imagedescription.

Cross-parametric generative adversarial network-based magnetic resonance image feature synthesis for breast lesion classification.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) contains information on tumor morphology and physiology for breast cancer diagnosis and treatment. However, this technology requires contrast agent injection with more acquisition time than other parametric images, such as T2-weighted imaging (T2WI). Current image synthesis methods attempt to map the image data from one domain to another, whereas it is challenging or even infeasible to map the images with one sequence into images with multiple sequences. Here, we propose a new approach of cross-parametric generative adversarial network (GAN)-based feature synthesis (CPGANFS) to generate discriminative DCE-MRI features from T2WI with applications in breast cancer diagnosis. The proposed approach decodes the T2W images into latent cross-parameter features to reconstruct the DCE-MRI and T2WI features by balancing the information shared between the two. A Wasserstein GAN with a gradient penalty is employed to differentiate the T2WI-generated features from ground-truth features extracted from DCE-MRI. The synthesized DCE-MRI feature-based model achieved significantly (p = 0.036) higher prediction performance (AUC = 0.866) in breast cancer diagnosis than that based on T2WI (AUC = 0.815). Visualization of the model shows that our CPGANFS method enhances the predictive power by levitating attention to the lesion and the surrounding parenchyma areas, which is driven by the interparametric information learned from T2WI and DCE-MRI. Our proposed CPGANFS provides a framework for cross-parametric MR image feature generation from a single-sequence image guided by an information-rich, time-series image with kinetic information. Extensive experimental results demonstrate its effectiveness with high interpretability and improved performance in breast cancer diagnosis.

Synthetic MRI in breast cancer: differentiating benign from malignant lesions and predicting immunohistochemical expression status

To evaluate and compare the performance of synthetic magnetic resonance imaging (SyMRI) in classifying benign and malignant breast lesions and predicting the expression status of immunohistochemistry (IHC) markers. We retrospectively analysed 121 patients with breast lesions who underwent dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and SyMRI before surgery in our hospital. DCE-MRI was used to assess the lesions, and then regions of interest (ROIs) were outlined on SyMRI (before and after enhancement), and apparent diffusion coefficient (ADC) maps to obtain quantitative values. After being grouped according to benign and malignant status, the malignant lesions were divided into high and low expression groups according to the expression status of IHC markers. Logistic regression was used to analyse the differences in independent variables between groups. The performance of the modalities in classification and prediction was evaluated by receiver operating characteristic (ROC) curves. In total, 57 of 121 lesions were benign, the other 64 were malignant, and 56 malignant lesions performed immunohistochemical staining. Quantitative values from proton density-weighted imaging prior to an injection of the contrast agent (PD-Pre) and T2-weighted imaging (T2WI) after the injection (T2-Gd), as well as its standard deviation (SD of T2-Gd), were valuable SyMRI parameters for the classification of benign and malignant breast lesions, but the performance of SyMRI (area under the curve, AUC = 0.716) was not as good as that of ADC values (AUC = 0.853). However, ADC values could not predict the expression status of breast cancer markers, for which SyMRI had excellent performance. The AUCs of androgen receptor (AR), estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER-2), p53 and Ki-67 were 0.687, 0.890, 0.852, 0.746, 0.813 and 0.774, respectively. SyMRI had certain value in distinguishing between benign and malignant breast lesions, and ADC values were still the ideal method. However, to predict the expression status of IHC markers, SyMRI had an incomparable value compared with ADC values.

Imaging classification and BIRADS assessment of cystic breast lesions with pathologic correlates. a 5-year experience in Zaria, North West Nigeria.

Breast cysts encompass a variety of pathologies, both benign and malignant. Therefore, classifying cysts into different categories is needful to develop a management algorithm. This study aimed to describe and distinguish between simple, complicated and complex cysts; and compare the final BIRADS assessment with pathologic findings. A 5-year retrospective review of our ultrasound database identified two hundred and seventy patients with cystic breast lesions. They were divided into simple, complicated and complex cysts according to ultrasound characteristics based on shape, orientation, margin, wall thickness, internal features (echogenicity, septa, mass) posterior acoustic features, surrounding tissue vascularity. The final BIRADS assessment was correlated with histological findings. There were two hundred and sixty-six (98.5%) females and four (1.5%) males with a mean age 34.9 ± 11.8 years. The commonest presentation was a palpable mass, in 70% of the patients. There were 89 (33.0%) simple cysts, 61 (22.6%) complicated cysts and 120 (44.4%) complex cysts. Majority of the breast cysts (83%) were benign with overall 17% incidence of malignancy. Complex cysts were the most frequent cyst type in our study, it is also the category most frequently associated with breast cancer, obviating the need for histology.

Breast lesions segmentation and classification in a two-stage process based on Mask-RCNN and Transfer Learning

Breast lesions segmentation and classification in a two-stage process based on Mask-RCNN and Transfer Learning

Classification of breast lesions in ultrasound images using deep convolutional neural networks: transfer learning versus automatic architecture design

Deep convolutional neural networks (DCNNs) have demonstrated promising performance in classifying breast lesions in 2D ultrasound (US) images. Exiting approaches typically use pre-trained models based on architectures designed for natural images with transfer learning. Fewer attempts have been made to design customized architectures specifically for this purpose. This paper presents a comprehensive evaluation on transfer learning based solutions and automatically designed networks, analyzing the accuracy and robustness of different recognition models in three folds. First, we develop six different DCNN models (BNet, GNet, SqNet, DsNet, RsNet, IncReNet) based on transfer learning. Second, we adapt the Bayesian optimization method to optimize a CNN network (BONet) for classifying breast lesions. A retrospective dataset of 3034 US images collected from various hospitals is then used for evaluation. Extensive tests show that the BONet outperforms other models, exhibiting higher accuracy (83.33%), lower generalization gap (1.85%), shorter training time (66 min), and less model complexity (approximately 0.5 million weight parameters). We also compare the diagnostic performance of all models against that by three experienced radiologists. Finally, we explore the use of saliency maps to explain the classification decisions made by different models. Our investigation shows that saliency maps can assist in comprehending the classification decisions.Graphical

Future Practices of Breast Pathology Using Digital and Computational Pathology.

Pathology clinical practice has evolved by adopting technological advancements initially regarded as potentially disruptive, such as electron microscopy, immunohistochemistry, and genomic sequencing. Breast pathology has a critical role as a medical domain, where the patient's pathology diagnosis has significant implications for prognostication and treatment of diseases. The advent of digital and computational pathology has brought about significant advancements in the field, offering new possibilities for enhancing diagnostic accuracy and improving patient care. Digital slide scanning enables to conversion of glass slides into high-fidelity digital images, supporting the review of cases in a digital workflow. Digitization offers the capability to render specimen diagnoses, digital archival of patient specimens, collaboration, and telepathology. Integration of image analysis and machine learning-based systems layered atop the high-resolution digital images offers novel workflows to assist breast pathologists in their clinical, educational, and research endeavors. Decision support tools may improve the detection and classification of breast lesions and the quantification of immunohistochemical studies. Computational biomarkers may help to contribute to patient management or outcomes. Furthermore, using digital and computational pathology may increase standardization and quality assurance, especially in areas with high interobserver variability. This review explores the current landscape and possible future applications of digital and computational techniques in the field of breast pathology.

Optimizing proportional balance between supervised and unsupervised features for ultrasound breast lesion classification

Computer-aided-diagnosis systems have led to transformative advancements to breast ultrasound interpretations, surmounting the challenges of manual assessments. Various methods have been explored for precisely classifying breast nodules. Notably, the channel embedding approach and the integration of characteristics from different modalities have garnered significant attention thus far. However, a serious drawback of existing methodologies lies in their reliance on mask availability or the necessity for complex preprocessing procedures to extract principal features.In this paper, we introduce a simple yet powerful lesion characterization framework that effectively addresses these shortcomings while achieving satisfactory discrimination performance. Our method simultaneously employs independent networks, trained via both supervised and unsupervised learning. Specifically, three supervised backbones are selected where the final feature volumes are progressively doubled in scale. Moreover, convolutional autoencoders are developed to perform unsupervised representation learning. By adjusting its hyperparameters, we generate multiple latent variables that align with the feature volumes of the supervised networks. Hence, our primary objective is to analyze an effective hybridization scheme driven by diverse automated representations and validate the optimized proportion of each component.Extensive ratio experiments are conducted on two public breast ultrasound datasets. The quantitative results highlight the importance of attaining a proper feature balance in hybridization. Further, consistent observations revealed the optimal ratio of 1:2 between the supervised and unsupervised volumes. We report that this condition derived a remarkable enhancement in discriminating tumor categories. To verify its robustness, our method is compared with recent competing approaches and demonstrated superiority without the need for additional specific requirements.