Breast Ultrasound Image Classification Research Articles

Revolutionizing breast ultrasound diagnostics with EfficientNet-B7 and Explainable AI

Breast cancer is a leading cause of mortality among women globally, necessitating precise classification of breast ultrasound images for early diagnosis and treatment. Traditional methods using CNN architectures such as VGG, ResNet, and DenseNet, though somewhat effective, often struggle with class imbalances and subtle texture variations, leading to reduced accuracy for minority classes such as malignant tumors. To address these issues, we propose a methodology that leverages EfficientNet-B7, a scalable CNN architecture, combined with advanced data augmentation techniques to enhance minority class representation and improve model robustness. Our approach involves fine-tuning EfficientNet-B7 on the BUSI dataset, implementing RandomHorizontalFlip, RandomRotation, and ColorJitter to balance the dataset and improve model robustness. The training process includes early stopping to prevent overfitting and optimize performance metrics. Additionally, we integrate Explainable AI (XAI) techniques, such as Grad-CAM, to enhance the interpretability and transparency of the model’s predictions, providing visual and quantitative insights into the features and regions of ultrasound images influencing classification outcomes. Our model achieves a classification accuracy of 99.14%, significantly outperforming existing CNN-based approaches in breast ultrasound image classification. The incorporation of XAI techniques enhances our understanding of the model’s decision-making process, thereby increasing its reliability and facilitating clinical adoption. This comprehensive framework offers a robust and interpretable tool for the early detection and diagnosis of breast cancer, advancing the capabilities of automated diagnostic systems and supporting clinical decision-making processes.

An Interpretable Breast Ultrasound Image Classification Algorithm Based on Convolutional Neural Network and Transformer

An Interpretable Breast Ultrasound Image Classification Algorithm Based on Convolutional Neural Network and Transformer

Gradually Vanishing Bridge Based on Multi-Kernel Maximum Mean Discrepancy for Breast Ultrasound Image Classification

Gradually Vanishing Bridge Based on Multi-Kernel Maximum Mean Discrepancy for Breast Ultrasound Image Classification

Deep learning and genetic algorithm-based ensemble model for feature selection and classification of breast ultrasound images

Feature extraction and selection are important techniques in the classification of medical images. Extraction of key features and selection of relevant features are the preliminary processes that are essential for identifying the shape of an object or diagnosis of a tumor in images. In this study, we conduct a thorough comparison of deep neural networks' performance. The comparison of the networks infers MobileNet as optimal for feature extraction from medical images, where it has minimal parameters and high validation accuracy. For feature selection, we employ the Genetic Algorithm (GA) because of its proficiency in handling high-dimensional and complex feature space. GA's iterative process aligns well with the unique characteristics of breast ultrasound (BUS) images, which enhances its efficacy in selecting salient features of BUS images. An ensemble model, capitalizing on the collective decision-making capabilities of multiple classifiers based on a weighted voting scheme for classification, is proposed. Empirical evaluations are conducted using two publicly available BUS image datasets, BUSI and UDIAT. The proposed model demonstrates a notable improvement of approximately 4% and 9% in accuracy for the BUSI and UDIAT, respectively. The improved diagnostic accuracy in breast abnormality identification allows for early abnormality diagnosis. This improves treatment outcomes for breast cancer patients and highlights the practical value of the proposed method for improving BUS image categorization.

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.

Distilling Knowledge From an Ensemble of Vision Transformers for Improved Classification of Breast Ultrasound

To develop a deep learning model for the automated classification of breast ultrasound images as benign or malignant. More specifically, the application of vision transformers, ensemble learning, and knowledge distillation is explored for breast ultrasound classification. Single view, B-mode ultrasound images were curated from the publicly available Breast Ultrasound Image (BUSI) dataset, which has categorical ground truth labels (benign vs malignant) assigned by radiologists and malignant cases confirmed by biopsy. The performance of vision transformers (ViT) is compared to convolutional neural networks (CNN), followed by a comparison between supervised, self-supervised, and randomly initialized ViT. Subsequently, the ensemble of 10 independently trained ViT, where the ensemble model is the unweighted average of the output of each individual model is compared to the performance of each ViT alone. Finally, we train a single ViT to emulate the ensembled ViT using knowledge distillation. On this dataset that was trained using five-fold cross validation, ViT outperforms CNN, while self-supervised ViT outperform supervised and randomly initialized ViT. The ensemble model achieves an area under the receiver operating characteristics curve (AuROC) and area under the precision recall curve (AuPRC) of 0.977 and 0.965 on the test set, outperforming the average AuROC and AuPRC of the independently trained ViTs (0.958±0.05 and 0.931±0.016). The distilled ViT achieves an AuROC and AuPRC of 0.972 and 0.960. Both transfer learning and ensemble learning can each offer increased performance independently and can be sequentially combined to collectively improve the performance of the final model. Furthermore, a single vision transformer can be trained to match the performance of an ensemble of a set of vision transformers using knowledge distillation.

Breast cancer prediction by ensemble meta-feature space generator based on deep neural network

BackgroundClassifying ultrasound images is one of the quickest approaches to diagnose breast cancer. In recent years, the medical community has increasingly turned to deep learning-based methods, which have demonstrated superior performance compared to other existing methods for most computer vision tasks. Since deep learning approaches suffer from over-fitting problem, the main challenge in the medical image analysis is the lack of enough available dataset to train deep learning models to extract efficient features. Many existing methods rely on data augmentation techniques to address the issue of over-fitting especially in imbalanced datasets. However, the process of finding appropriate data augmentation methods often involves trial and error, and it can be time-consuming. Therefore, there is a need to develop an algorithm that can automatically generate effective features, and minimize the reliance on a large amount of data to overcome over-fitting. MethodIn this study, we introduced a novel ensemble meta-feature space generator (EMFSG-Net) for the classification of breast ultrasound images. We first employed a transfer learning approach to obtain initial features from raw images. Then we applied our method to create a more efficient feature space from the initial features. Our method is inspired by ensemble methods that aim to minimize bias and variance errors. Ensemble methods are widely used in previous works based on classification models. The novelty of our approach is to adapt these techniques to regression-based models, which enables us to construct a meta-feature generator capable of producing an effective feature space. This method offers an effective solution to address the overfitting problem commonly encountered in deep learning for medical image processing. By obtaining more suitable features, it significantly improves the classification performance. Notably, this method eliminates the need for data augmentation, which saves time by avoiding the search for appropriate data augmentation techniques. This approach demonstrates its efficacy specifically in addressing imbalanced datasets, which are characterized by a disproportionate distribution of class labels. ResultWe have evaluated our model on BUSI dataset, a collection of breast ultrasound images that is widely used in scientific researches to detect breast cancer. We have achieved the accuracy of 97.96% and F1-score of 96.2% on the dataset. ConclusionThe experimental results show that the proposed model has increased 4.8% of the classification F1-score compared to other feature extraction and deep learning methods.

Breast ultrasound image classification and physiological assessment based on GoogLeNet

BackgroundMedical ultrasound image classification based on convolutional neural network is the mainstream breast cancer classification model, but its limited perceptual ability limits its ability to obtain global information. MethodologyA total of 880 breast ultrasound images were collected from 700 patients including 103 normal images, 467 malignant tumor images and 210 benign tumor images. In this paper, the diagnosis of breast ultrasound images was realized by constructing the CNN model of GoogLeNet. Firstly, the breast images were preprocessed based on the TV model. After that, the CNN model was trained and a more accurate model with a wider range of application was obtained based on improved Inception. Then we extract image features of different sizes; Then, feature classification was completed in different model classifiers to realize benign and malignant detection of breast cancer. Meanwhile, comparative analysis was performed to verify the excellence of the GoogLeNet model. ResultsThe training time for classification was effectively reduced, and the classification accuracy rate was improved, reaching 96.37% combined with transformer learning. The loss value is down to 0.3492. Then, the classification accuracy of different structural models is discussed in the two models. The results show that GoogLeNet has great advantages in the detection of breast cancer ultrasound images. The influence of migration on experimental results is further discussed. Finally, combined with transfer learning, the three models were tested separately. The results show that transfer learning can improve system performance. ConclusionBased on the TV model and the GoogLeNet model, this paper designs a method of breast cancer detection and classification based on the combination of TV and GoogLeNet model and transfer learning. Experiments show that this method can repair part of the texture damaged by markers in ultrasonic images, effectively extract the features of thyroid nodule, and accurately judge whether the breast is diseased, which greatly improves the diagnostic efficiency of doctors.

Breast ultrasound image classification with hard sample generation and semi-supervised learning

Ultrasound is a fast and non-invasive imaging technique widely used in the early-stage diagnosis of breast cancer. Ultrasound images are also commonly used in deep learning, but traditional computer algorithms and even ultrasound diagnosticians often struggle to accurately distinguish between benign and malignant cases, as some ultrasound images exhibit mixed characteristics of both. To address the above problems, we defined hard sample mining rules firstly. To address the above problems, we defined rules for hard sample mining and utilized shape descriptors (Concavity, Growth Direction, Compactness, Circle Variance, and Elliptic Variance) to describe the samples. For those samples with overlapped features of both categories, i.e., benign and malignant, we identify them as hard samples. In order to make the network pay more attention to hard samples, we propose a Phased GAN (PGAN) that can generate hard samples to increase the number of such samples in the dataset. However, since the generated hard samples lack labels, they cannot be used in supervised training. To overcome this limitation, we propose a new semi-supervised network (SSN) that trains them with labeled samples. We also add a hard sample loss function to the SSN to regularize the features of the hard samples, which helps the network learn these features better. In summary, we propose a breast ultrasound image classification using PGAN and SSN, which achieves a classification accuracy of 97.7% on a public breast dataset (445 cases of benign and 210 cases of malignant).

Breast UltraSound Image classification using fuzzy-rank-based ensemble network

Breast Cancer is one of the most commonly occurring cancers in women. Detecting cancer at an earlier stage increases the chance of survival. Thus the development of an automatic CAD system is important for cancer diagnosis. Researchers have used various image modalities like mammograms and Magnetic Resonance Images to detect cancer. This manuscript uses UltraSound Images for detecting breast cancer. We have proposed a fuzzy-rank-based ensemble network for detecting breast cancer. The proposed model uses four different base learners, namely VGG-Net, DenseNet, Xception, and Inception, and it tries to take advantage of the predictions made by the base learners. The weights of the initial layers of the base learners are pre-trained on the ImageNet dataset. In contrast, the final five layers are fine-tuned using a publicly available Breast Ulta Sound Image dataset. The fuzzy rank of the base learners’ predictions is used to make the final classification. Conducting five-fold cross-validation using the base learners an accuracy of 77.69±3.22, 83.23±3.14, 78.31±2.27, and 78.62±4.23 were obtained. Furthermore, using the proposed fuzzy-rank-based model, an accuracy of 85.23±2.52 is obtained. We have proved that the proposed fuzzy-rank-based ensemble network increases the classification performance.

Quantitative feature classification for breast ultrasound images using improved naive bayes

AbstractThe quantitative feature classification of breast ultrasound images can explore the intrinsic connection between image features and lesions, and become an important basis for the diagnosis of tumor nature in clinical medicine. However, the tumor features extracted from breast ultrasound images are generally based on the overall situation, which has an impact on the classification of features in ultrasound images. Therefore, a quantitative feature classification algorithm for breast ultrasound images using improved Naive Bayes (NB) is proposed. First, histogram features, grayscale co‐generation matrix and other texture features are extracted from the ultrasound image. Second, NB is combined with decision tree calculation to improve the traditional NB algorithm. Finally, feature classification is optimized using improved NB algorithm to achieve high accuracy in the classification of ultrasound images. The results show that the contour of the lesion recognition results obtained by the proposed algorithm is the most complete, and the other regions can maximize the retention of the texture features of the original image. The classification accuracy is high, and the classification time of quantitative features of breast ultrasound images is short. It has application value in the field of breast ultrasound diagnosis.

Deep Learning Approaches for Classification of Breast Cancer in Ultrasound (US) Images

Breast cancer is one of the deadliest cancer types affecting women worldwide. As with all types of cancer, early detection of breast cancer is of vital importance. Early diagnosis plays an important role in reducing deaths and fighting cancer. Ultrasound (US) imaging is a painless and common technique used in the early detection of breast cancer. In this article, deep learning-based approaches for the classification of breast US images have been extensively reviewed. Classification performance of breast US images of architectures such as AlexNet, VGG, ResNet, GoogleNet and EfficientNet, which are among the most basic CNN architectures, has been compared. Then, transformer models, which are one of the most popular deep learning architectures these days and show similar performance to the performance of CNN' architectures in medical images, are examined. BUSI, the only publicly available dataset, was used in experimental studies. Experimental studies have shown that the transformer and CNN models successfully classify US images of the breast. It has been observed that vision transformer model outperforms other models with 88.6% accuracy, 90.1% precison, 87.4% recall and 88.7% F1-score. This study shows that deep learning architectures are successful in classification of US images and can be used in the clinic experiments in the near future.

Automated BI-RADS classification of lesions using pyramid triple deep feature generator technique on breast ultrasound images.

Ultrasound (US) is an important imaging modality used to assess breast lesions for malignant features. In the past decade, many machine learning models have been developed for automated discrimination of breast cancer versus normal on US images, but few have classified the images based on the Breast Imaging Reporting and Data System (BI-RADS) classes. This work aimed to develop a model for classifying US breast lesions using a BI-RADS classification framework with a new multi-class US image dataset. We proposed a deep model that combined a novel pyramid triple deep feature generator (PTDFG) with transfer learning based on three pre-trained networks for creating deep features. Bilinear interpolation was applied to decompose the input image into four images of successively smaller dimensions, constituting a four-level pyramid for downstream feature generation with the pre-trained networks. Neighborhood component analysis was applied to the generated features to select each network's 1,000 most informative features, which were fed to support vector machine classifier for automated classification using a ten-fold cross-validation strategy. Our proposed model was validated using a new US image dataset containing 1,038 images divided into eight BI-RADS classes and histopathological results. We defined three classification schemes: Case 1 involved the classification of all images into eight categories; Case 2, classification of breast US images into five BI-RADS classes; and Case 3, classification of BI-RADS 4 lesions into benign versus malignant classes. Our PTDFG-based transfer learning model attained accuracy rates of 79.29%, 80.42%, and 88.67% for Case 1, Case 2, and Case 3, respectively.

Classification of Breast Ultrasound Images: An Analysis Using Machine Intelligent Based Approach

Classification of Breast Ultrasound Images: An Analysis Using Machine Intelligent Based Approach

Classification of Breast Ultrasound Images: An Analysis Using Machine Intelligent Based Approach

Purpose: Breast Cancer (BC) is considered as one of the most dangerous diseases, especially in women. The survivability of the patient is a challenging task if the breast cancer is in severe stage. It is very much essential for the early classification of breast ultrasound images (BUIs) into several categories such as benign (BN), malignant (MG) and normal (NL), etc. so that preventive measures can be taken accordingly at the earliest. Approach: In this work, a machine intelligent (MI) based approach is proposed for the classification of BUIs into the BN, MG and NL types. The proposed approach is focused on the stacking (hybridization) of Logistic Regression (LRG), Support Vector Machine (SVMN), Random Forest (RFS) and Neural Network (NNT) methods to carry out such classification. The proposed method is compared with other machine learning (ML) based methods such as LRG, SVMN, RFS, NNT, Decision Tree (DTR), AdaBoost (ADB), Naïve Bayes (NBY), K-Nearest Neighbor (KNNH) and Stochastic Gradient Descent (SGDC) for performance analysis. Result: The proposed method and other ML based methods have been implemented using Python based Orange 3.26.0. In this work, 750 TLDIs having 250 numbers of each type such as BN, MG and NL are taken from the Kaggle source. The performance of all the methods is assessed using the performance parameters such as classification accuracy (CA), F1, Precision (PR) and Recall (RC). From the results, it is found that the proposed method is capable of providing better classification results in terms of CA, F1, PR and RC as compared to other ML based methods such as LRG, SVMN, RFS, NNT, DTR, ADB, NBY, KNNH and SGD. Originality: In this work, a MI based approach is proposed by focusing on the stacking of LRG, SVMN, RFS and NNT methods to carry out the classification of BUIs into several types such as BN, MG and NL. The proposed approach performs better in terms of CA, F1, PR and RC as compared to LRG, SVMN, RFS, NNT, DTR, ADB, NBY, KNNH and SGDC methods. Paper Type: Conceptual Research.

Fully-automated deep learning pipeline for segmentation and classification of breast ultrasound images

Breast cancer is the most prevalent type of cancer among the female world population. Its early detection has a crucial role in enhancing the effectiveness of treatments, as well as reducing serious complications and deaths. Ultrasound imaging represents a standard diagnostic technique for this purpose, due to its low invasiveness and cost. However, as this technique is susceptible to a certain degree of uncertainty, computer-aided solutions have been proposed in recent years to reduce the operator workload and improve the accuracy of diagnoses. Following this trend, the present study aims to design and propose a fully-automated and multi-layer pipeline for the segmentation and classification of breast lesions associated with cancer risk, from ultrasound images. To achieve this goal, we first evaluate and compare the performance of several Convolutional Neural Network (CNN) architectures in tackling the above tasks. Then, we combine the performance of such networks through specialized ensembles, to better discriminate among heterogeneous cases. Lastly, we present a novel step of cyclic mutual optimization that exploits the intermediate results of the classification step to improve the segmentation outcome and vice versa, in an iterative manner. Experimental findings obtained on public datasets show the superiority of the ensemble methods over the individual networks. Moreover, with a Dice coefficient of ∼82% in the segmentation task and an accuracy of ∼91% in the classification task, our best configuration also shows to be competitive with respect to the state-of-the-art.

Vision Transformers for Classification of Breast Ultrasound Images.

Medical ultrasound (US) imaging has become a prominent modality for breast cancer imaging due to its ease of use, low cost, and safety. In the past decade, convolutional neural networks (CNNs) have emerged as the method of choice in vision applications and have shown excellent potential in the automatic classification of US images. Despite their success, their restricted local receptive field limits their ability to learn global context information. Recently, Vision Transformer (ViT) designs, based on self-attention between image patches, have shown great potential to be an alternative to CNNs. In this study, for the first time, we utilize ViT to classify breast US images using different augmentation strategies. We also adopted a weighted cross-entropy loss function since breast ultrasound datasets are often imbalanced. The results are provided as classification accuracy and Area Under the Curve (AUC) metrics, and the performance is compared with the SOTA CNNs. The results indicate that the ViT models have comparable efficiency with or even better than the CNNs in the classification of US breast images. Clinical relevance- This work shows the potential of Vision Transformers in the automatic classification of masses in breast ultrasound, which helps clinicians diagnose and make treatment decisions more precisely.

Dynamic adversarial domain adaptation based on multikernel maximum mean discrepancy for breast ultrasound image classification

Due to the privacy of breast ultrasound images, it is difficult to obtain a dataset, which results in the lack of a large number of labeled datasets. Additionally, the difference in the feature distribution of datasets is also considered. In this paper, a dynamic adversarial domain adaptive network based on the multi kernel maximum mean discrepancy (MK_DAAN) is proposed. An adaptive layer is added to the model to further align the feature distribution of the datasets in the source and target domains, and the multi kernel maximum mean discrepancy is adopted in the adaptive distance measurement. The dual feature alignment of the adaptive layer and adversarial learning improves the classification performance of the model in breast ultrasound images. The experimental results show that the MK_DAAN model has better classification performance than typical unsupervised domain adaptation (DDC, DAN) and typical adjustment domain adaptation (DAAN, MADA) models.

Comprehensive Study for Breast Cancer Using Deep Learning and Traditional Machine Learning

Comprehensive Study for Breast Cancer Using Deep Learning and Traditional Machine Learning

LRSCnet: Local Reference Semantic Code learning for breast tumor classification in ultrasound images

PurposeThis study proposed a novel Local Reference Semantic Code (LRSC) network for automatic breast ultrasound image classification with few labeled data.MethodsIn the proposed network, the local structure extractor is firstly developed to learn the local reference which describes common local characteristics of tumors. After that, a two-stage hierarchical encoder is developed to encode the local structures of lesion into the high-level semantic code. Based on the learned semantic code, the self-matching layer is proposed for the final classification.ResultsIn the experiment, the proposed method outperformed traditional classification methods and AUC (Area Under Curve), ACC (Accuracy), Sen (Sensitivity), Spec (Specificity), PPV (Positive Predictive Values), and NPV(Negative Predictive Values) are 0.9540, 0.9776, 0.9629, 0.93, 0.9774 and 0.9090, respectively. In addition, the proposed method also improved matching speed.ConclusionsLRSC-network is proposed for breast ultrasound images classification with few labeled data. In the proposed network, a two-stage hierarchical encoder is introduced to learn high-level semantic code. The learned code contains more effective high-level classification information and is simpler, leading to better generalization ability.