Machine Learning for Workflow Applications in Screening Mammography: Systematic Review and Meta-Analysis

Abstract

EVIDENCE-BASED PRACTICET here are now more than five U.S. Food and Drug Administration-approved algorithms for mammographic interpretation, primarily to be used as clinical decision support systems (1). Research has demonstrated that these machine learning (ML) computer-aided detection (CAD) algorithms can reach and even exceed clinician performance, providing an independent definitive output (ie, case-level decision) on two-dimensional standard-view mammogram (ie, mediolateral oblique and craniocaudal) data (Fig 1) (2,3). This could allow for ML stand-alone CAD and computer-aided diagnosis (CADx), or, when ML algorithms are set at a high sensitivity, for the automated case-based computer-aided triage (CADt) of mammograms within the screen reading workflow (4).Many countries have implemented breast screening to detect cancer at an earlier stage, albeit with differing screening processes, such as single reading in the United States and double reading in many European countries, with screening starting at varied ages (40-50 years) and differing intervals between screening (annual, biennial, and triennial) (5-8). Mammography remains the most common imaging modality used, although its cost-effectiveness is debated because of false-positive findings, overdiagnosis, and false-negative findings (ie, interval cancers) (9,10). Human readers-for example, radiologists and reporting radiographers in the United Kingdom-are under increasing pressure because of increasing workloads, demands from busy clinics, strict Background: Advances in computer processing and improvements in data availability have led to the development of machine learning (ML) techniques for mammographic imaging.Purpose: To evaluate the reported performance of stand-alone ML applications for screening mammography workflow. Materials and Methods:Ovid Embase, Ovid Medline, Cochrane Central Register of Controlled Trials, Scopus, and Web of Science literature databases were searched for relevant studies published from January 2012 to September 2020. The study was registered with the PROSPERO International Prospective Register of Systematic Reviews (protocol no. CRD42019156016). Stand-alone technology was defined as a ML algorithm that can be used independently of a human reader. Studies were quality assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 and the Prediction Model Risk of Bias Assessment Tool, and reporting was evaluated using the Checklist for Artificial Intelligence in Medical Imaging. A primary meta-analysis included the top-performing algorithm and corresponding reader performance from which pooled summary estimates for the area under the receiver operating characteristic curve (AUC) were calculated using a bivariate model.Results: Fourteen articles were included, which detailed 15 studies for stand-alone detection (n = 8) and triage (n = 7). Triage studies reported that 17%-91% of normal mammograms identified could be read by adapted screening, while "missing" an estimated 0%-7% of cancers. In total, an estimated 185 252 case...

A Novel CNN-Inception-V4-Based Hybrid Approach for Classification of Breast Cancer in Mammogram Images

“…X-ray scans of each breast are acquired from two perspectives during a mammography screening. Human readers, such as an expert radiologist, assess and examine screening mammograms [6]. Dual reading was discovered to enhance evaluation results and has been implemented in many countries.…”

Stand-Alone Machine Learning: More Work Is Needed

“…CADx uses artificial intelligence-based quantitative features, and it is based on deep learning. CADt systems serve as the first reader, and these systems can be configured to prioritize the worklist for the radiologist who serves as the second reader (1).…”

“…The performance of screening mammography algorithms in stand-alone CAD and CADx tasks was found to be almost as good as that of human readers, according to the researchers who conducted the meta-analysis (1). After conducting a comparison with two recently published reader studies, Hickman et al (1) discovered that although the pooled sensitivity of the algorithms (75.4%) was higher than that of pooled readers (73.0%) and single reading in Sweden (73.0%), it was inferior to single reading in the United States (86.9%) and to double reading with consensus in Sweden (85.0%) (1,6,7). The pooled specificity of the ML algorithms (90.6%) was superior to that of the pooled readers (88.6%) and single reading in the United States (88.9%), but it was inferior to single (96.0%) and double (98.0%) reading with consensus in Sweden (1,6,7).…”

“…The pooled specificity of the ML algorithms (90.6%) was superior to that of the pooled readers (88.6%) and single reading in the United States (88.9%), but it was inferior to single (96.0%) and double (98.0%) reading with consensus in Sweden (1,6,7). The bottom line, according to the authors, is that the Stand-Alone Machine Learning findings of the meta-analysis demonstrate that further improvements in ML learning systems are required to achieve clinically relevant performance targets in screening mammography (1). ML can process and triage larger case volumes at a faster rate than humans, allowing for more efficient case management.…”

“…Ideally, in the future, ML algorithms should be trained and refined using very large, high-quality data sets that are unbiased and generalizable, along with well-established ground truth measures (8). In one of the studies reviewed by Hickman et al (1), a mammography study was evaluated using US results as a ground truth. This is an unacceptable method of verifying the presence or absence of malignancy in a mammography study.…”