Global detection approach for clustered microcalcifications in mammograms using a deep learning network

Reproducibility and Explainability of Deep Learning in Mammography: A Systematic Review of Literature

Background  Although abundant literature is currently available on the use of deep learning for breast cancer detection in mammography, the quality of such literature is widely variable. Purpose  To evaluate published literature on breast cancer detection in mammography for reproducibility and to ascertain best practices for model design. Methods  The PubMed and Scopus databases were searched to identify records that described the use of deep learning to detect lesions or classify images into cancer or noncancer. A modification of Quality Assessment of Diagnostic Accuracy Studies (mQUADAS-2) tool was developed for this review and was applied to the included studies. Results of reported studies (area under curve [AUC] of receiver operator curve [ROC] curve, sensitivity, specificity) were recorded. Results  A total of 12,123 records were screened, of which 107 fit the inclusion criteria. Training and test datasets, key idea behind model architecture, and results were recorded for these studies. Based on mQUADAS-2 assessment, 103 studies had high risk of bias due to nonrepresentative patient selection. Four studies were of adequate quality, of which three trained their own model, and one used a commercial network. Ensemble models were used in two of these. Common strategies used for model training included patch classifiers, image classification networks (ResNet in 67%), and object detection networks (RetinaNet in 67%). The highest reported AUC was 0.927 ± 0.008 on a screening dataset, while it reached 0.945 (0.919-0.968) on an enriched subset. Higher values of AUC (0.955) and specificity (98.5%) were reached when combined radiologist and Artificial Intelligence readings were used than either of them alone. None of the studies provided explainability beyond localization accuracy. None of the studies have studied interaction between AI and radiologist in a real world setting. Conclusion  While deep learning holds much promise in mammography interpretation, evaluation in a reproducible clinical setting and explainable networks are the need of the hour.

Computer-aided breast cancer detection and classification in mammography: A comprehensive review

Cancer is the second cause of mortality worldwide and it has been identified as a perilous disease. Breast cancer accounts for ∼20% of all new cancer cases worldwide, making it a major cause of morbidity and mortality. Mammography is an effective screening tool for the early detection and management of breast cancer. However, the identification and interpretation of breast lesions is challenging even for expert radiologists. For that reason, several Computer-Aided Diagnosis (CAD) systems are being developed to assist radiologists to accurately detect and/or classify breast cancer. This review examines the recent literature on the automatic detection and/or classification of breast cancer in mammograms, using both conventional feature-based machine learning and deep learning algorithms. The review begins with a comparison of algorithms developed specifically for the detection and/or classification of two types of breast abnormalities, micro-calcifications and masses, followed by the use of sequential mammograms for improving the performance of the algorithms. The available Food and Drug Administration (FDA) approved CAD systems related to triage and diagnosis of breast cancer in mammograms are subsequently presented. Finally, a description of the open access mammography datasets is provided and the potential opportunities for future work in this field are highlighted. The comprehensive review provided here can serve both as a thorough introduction to the field but also provide indicative directions to guide future applications.

Anomaly Detection of Calcifications in Mammography Based on 11,000 Negative Cases

In mammography, calcifications are one of the most common signs of breast cancer. Detection of such lesions is an active area of research for computer-aided diagnosis and machine learning algorithms. Due to limited numbers of positive cases, many supervised detection models suffer from overfitting and fail to generalize. We present a one-class, semi-supervised framework using a deep convolutional autoencoder trained with over 50,000 images from 11,000 negative-only cases. Since the model learned from only normal breast parenchymal features, calcifications produced large signals when comparing the residuals between input and reconstruction output images. As a key advancement, a structural dissimilarity index was used to suppress non-structural noises. Our selected model achieved pixel-based AUROC of 0.959 and AUPRC of 0.676 during validation, where calcification masks were defined in a semi-automated process. Although not trained directly on any cancers, detection performance of calcification lesions on 1,883 testing images (645 malignant and 1238 negative) achieved 75% sensitivity at 2.5 false positives per image. Performance plateaued early when trained with only a fraction of the cases, and greater model complexity or a larger dataset did not improve performance. This study demonstrates the potential of this anomaly detection approach to detect mammographic calcifications in a semi-supervised manner with efficient use of a small number of labeled images, and may facilitate new clinical applications such as computer-aided triage and quality improvement.

Automatic Pectoral Muscle Removal and Microcalcification Localization in Digital Mammograms

Objectives

Breast cancer is the most common cancer diagnosed in women, and microcalcification (MCC) clusters act as an early indicator. Thus, the detection of MCCs plays an important role in diagnosing breast cancer.

Methods

This paper presents a methodology for mammogram preprocessing and MCC detection. The preprocessing method employs automatic artefact deletion and pectoral muscle removal based on region-growing segmentation and polynomial contour fitting. The MCC detection method uses a convolutional neural network for region-of-interest (ROI) classification, along with morphological operations and wavelet reconstruction to reduce false positives (FPs).

Results

The methodology was evaluated using the mini-MIAS and UTP datasets in terms of segmentation accuracy in the preprocessing phase, as well as sensitivity and the mean FP rate per image in the MCC detection phase. With the mini-MIAS dataset, the proposed methods achieved accuracy scores of 99% for breast segmentation and 95% for pectoral segmentation, a sensitivity score of 82% for MCC detection, and an FP rate per image of 3.27. With the UTP dataset, the methods achieved accuracy scores of 97% for breast segmentation and 91% for pectoral segmentation, a sensitivity score of 78% for MCC detection, and an FP rate per image of 0.74.

Conclusions

The proposed preprocessing method outperformed the state-of-the-art methods for breast segmentation and achieved relatively good results for pectoral muscle removal. Furthermore, the MCC detection module achieved the highest test accuracy in identifying potential ROIs with MCCs compared to other methods.

Computer Vision-Based Microcalcification Detection in Digital Mammograms Using Fully Connected Depthwise Separable Convolutional Neural Network

Microcalcification clusters in mammograms are one of the major signs of breast cancer. However, the detection of microcalcifications from mammograms is a challenging task for radiologists due to their tiny size and scattered location inside a denser breast composition. Automatic CAD systems need to predict breast cancer at the early stages to support clinical work. The intercluster gap, noise between individual MCs, and individual object’s location can affect the classification performance, which may reduce the true-positive rate. In this study, we propose a computer-vision-based FC-DSCNN CAD system for the detection of microcalcification clusters from mammograms and classification into malignant and benign classes. The computer vision method automatically controls the noise and background color contrast and directly detects the MC object from mammograms, which increases the classification performance of the neural network. The breast cancer classification framework has four steps: image preprocessing and augmentation, RGB to grayscale channel transformation, microcalcification region segmentation, and MC ROI classification using FC-DSCNN to predict malignant and benign cases. The proposed method was evaluated on 3568 DDSM and 2885 PINUM mammogram images with automatic feature extraction, obtaining a score of 0.97 with a 2.35 and 0.99 true-positive ratio with 2.45 false positives per image, respectively. Experimental results demonstrated that the performance of the proposed method remains higher than the traditional and previous approaches.

Application of deep learning in the detection of breast lesions with four different breast densities

Objective

This retrospective study evaluated the model from populations with different breast densities and showed the model's performance on malignancy prediction.

Methods

A total of 608 mammograms were collected from Northern Jiangsu People's Hospital in Yangzhou City. The data from this province have not been used in the training or evaluation data set.

The model consists of three submodules, lesion detection (Mask‐rcnn), lesion registration between craniocaudal view and mediolateral oblique view, malignancy prediction network (ResNet). The data set used to train the model was obtained from nine institutions across six cities. For normal cases, there were no annotations. Here, we adopted the free‐response receiver operating characteristic (FROC) curve as the indicator to evaluate the detection performance of all cancers and triple‐negative breast cancer (TNBC). The FROC curves are also shown for mass/distortion/asymmetry and typical benign calcification in two kinds of populations with four types of breast density.

Results

The sensitivity to mass/distortion/asymmetry for the four types of breast (A, B, C, D) are 0.94, 0.92, 0.89, and 0.72, respectively, when false positive per image is 0.25, while these values are 1.00, 0.95, 0.92, and 0.90, respectively, for the amorphous calcification lesions. The sensitivity for the cancer is 0.85 at the same false‐positive rate. The TNBC accounts for about 10%–20% of all breast cancers and is more aggressive with poor prognosis than other breast cancers. Herein, we also evaluated performance on the TNBC cases. Our results show that Yizhun AI could detect 75% TNBC lesions at the same false‐positive level mentioned above.

Conclusion

The Yizhun AI model used in our work has good diagnostic efficiency for different types of breast, even for the extremely dense breast. It has a guiding role in the clinical diagnosis of breast cancer. The performance of Yizhun AI on mass/distortion/asymmetry is affected by breast density significantly compared to that on amorphous calcification.

Classification of Mammographic Breast Microcalcifications Using a Deep Convolutional Neural Network: A BI-RADS-Based Approach

MATERIALS AND METHODS

Over 56,000 images of 268 mammograms from 94 patients were labeled to 3 classes according to the BI-RADS standard: "no microcalcifications" (BI-RADS 1), "probably benign microcalcifications" (BI-RADS 2/3), and "suspicious microcalcifications" (BI-RADS 4/5). Using the preprocessed images, a dCNN was trained and validated, generating 3 types of models: BI-RADS 4 cohort, BI-RADS 5 cohort, and BI-RADS 4 + 5 cohort. For the final validation of the trained dCNN models, a test data set consisting of 141 images of 51 mammograms from 26 patients labeled according to the corresponding BI-RADS classification from the radiological reports was applied. The performances of the dCNN models were evaluated, classifying each of the mammograms and computing the accuracy in comparison to the classification from the radiological reports. For visualization, probability maps of the classification were generated.

RESULTS

The accuracy on the validation set after 130 epochs was 99.5% for the BI-RADS 4 cohort, 99.6% for the BI-RADS 5 cohort, and 98.1% for the BI-RADS 4 + 5 cohort. Confusion matrices of the "real-world" test data set for the 3 cohorts were generated where the radiological reports served as ground truth. The resulting accuracy was 39.0% for the BI-RADS 4 cohort, 80.9% for BI-RADS 5 cohort, and 76.6% for BI-RADS 4 + 5 cohort. The probability maps exhibited excellent image quality with correct classification of microcalcification distribution.

CONCLUSIONS

The dCNNs can be trained to successfully classify microcalcifications on mammograms according to the BI-RADS classification system in order to act as a standardized quality control tool providing the expertise of a team of radiologists.

Simultaneous Diagnosis of Severity and Features of Diabetic Retinopathy in Fundus Photography Using Deep Learning

Deep learning methods for diabetic retinopathy (DR) diagnosis are usually criticized as being lack of interpretability in the diagnostic result, thus limiting their application in clinic. Simultaneous prediction of DR related features during the DR severity diagnosis is able to resolve this issue by providing supporting evidence (i.e. DR related features) for the diagnostic result (i.e. DR severity). In this study, we propose a hierarchical multi-task deep learning framework for simultaneous diagnosis of DR severity and DR related features in fundus images. A hierarchical structure is introduced to incorporate the casual relationship between DR related features and DR severity levels. In the experiments, the proposed approach was evaluated on two independent testing sets using quadratic weighted Cohen's kappa coefficient, receiver operating characteristic analysis, and precision-recall analysis. A grader study was also conducted to compare the performance of the proposed approach with those of general ophthalmologists with different levels of experience. The results demonstrate that the proposed approach could improve the performance for both DR severity diagnosis and DR related feature detection when comparing with the traditional deep learning-based methods. It achieves performance close to general ophthalmologists with five years of experience when diagnosing DR severity levels, and general ophthalmologists with ten years of experience for referable DR detection.

Multiscale connected chain topological modelling for microcalcification classification

Computer-aided diagnosis (CAD) systems can be employed to help classify mammographic microcalcification clusters. In this paper, a novel method for the classification of the microcalcification clusters based on topology/connectivity has been introduced. The proposed method is distinct from existing techniques which concentrate on morphology and texture of microcalcifications and surrounding tissue. The proposed approach used multiscale morphological relationship of connectivity between microcalcifications where connected chains between nearest microcalcifications were generated at each scale. Subsequently, graph connectivity features at each scale were extracted to estimate the topological connectivity structure of microcalcification clusters for benign versus malignant classification. The proposed approach was evaluated using publicly available digitized datasets: MIAS and DDSM, in addition to the digital OPTIMAM dataset. The classification of features using KNN obtained a classification accuracy of 86.47±1.30%, 90.0±0.00%, 82.5±2.63%, 76.75±0.66% for the DDSM, MIAS-manual, MIAS-auto and OPTIMAM datasets respectively. The study showed that topological/connectivity modelling using a multiscale approach was appropriate for microcalcification cluster analysis and classification; topological connectivity and distribution can be linked to clinical understanding of microcalcification spatial distribution.

Microcalcification detection in full-field digital mammograms: A fully automated computer-aided system

BACKGROUND

Microcalcification clusters in mammograms can be considered as early signs of breast cancer. However, their detection is a very challenging task because of different factors: large variety of breast composition, highly textured breast anatomy, impalpable size of microcalcifications in some cases, as well as inherent low contrast of mammograms. Thus, the need to support the clinicians' work with an automatic tool.

METHODS

In this work a three-phases approach for clustered microcalcification detection is presented. Specifically, it is made up of a pre-processing step, aimed at highlighting potentially interesting breast structures, followed by a single microcalcification detection step, based on Hough transform, that is able to grasp the innate characteristic shape of the structures of interest. Finally, a cluster identification step to group microcalcifications is carried out by means of a clustering algorithm able to codify expert domain rules.

RESULTS

The detection performance of the proposed method has been evaluated on 364 mammograms of 182 patients obtaining a true positive ratio of 91.78% with 2.87 false positives per image.

CONCLUSIONS

Experimental results demonstrated that the proposed method is able to detect microcalcification clusters in digital mammograms showing performance comparable to different methodologies exploited in the state-of-art approaches, with the advantage that it does not require any training phase and a large set of data. The performance of the proposed approach remains high even for more difficult clinical cases of mammograms of young women having high-density breast tissue thus resulting in a reduced contrast between microcalcifications and surrounding dense tissues.

Convolutional Neural Networks for Radiologic Images: A Radiologist's Guide

Deep learning has rapidly advanced in various fields within the past few years and has recently gained particular attention in the radiology community. This article provides an introduction to deep learning technology and presents the stages that are entailed in the design process of deep learning radiology research. In addition, the article details the results of a survey of the application of deep learning-specifically, the application of convolutional neural networks-to radiologic imaging that was focused on the following five major system organs: chest, breast, brain, musculoskeletal system, and abdomen and pelvis. The survey of the studies is followed by a discussion about current challenges and future trends and their potential implications for radiology. This article may be used as a guide for radiologists planning research in the field of radiologic image analysis using convolutional neural networks.

A context-sensitive deep learning approach for microcalcification detection in mammograms

A challenging issue in computerized detection of clustered microcalcifications (MCs) is the frequent occurrence of false positives (FPs) caused by local image patterns that resemble MCs. We develop a context-sensitive deep neural network (DNN), aimed to take into account both the local image features of an MC and its surrounding tissue background, for MC detection. The DNN classifier is trained to automatically extract the relevant image features of an MC as well as its image context. The proposed approach was evaluated on a set of 292 mammograms using free-response receiver operating characteristic (FROC) analysis on the accuracy both in detecting individual MCs and in detecting MC clusters. The results demonstrate that the proposed approach could achieve significantly higher FROC curves when compared to two MC-based detectors. It indicates that incorporating image context information in MC detection can be beneficial for reducing the FPs in detections.