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

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 cases from three countries with more than 39 readers were included in the primary meta-analysis. The pooled sensitivity, specificity, and AUC was 75.4% (95% CI: 65.6, 83.2; P = .11), 90.6% (95% CI: 82.9, 95.0; P = .40), and 0.89 (95% CI: 0.84, 0.98), respectively, for algorithms, and 73.0% (95% CI: 60.7, 82.6), 88.6% (95% CI: 72.4, 95.8), and 0.85 (95% CI: 0.78, 0.97), respectively, for readers. Conclusion Machine learning (ML) algorithms that demonstrate a stand-alone application in mammographic screening workflows achieve or even exceed human reader detection performance and improve efficiency. However, this evidence is from a small number of retrospective studies. Therefore, further rigorous independent external prospective testing of ML algorithms to assess performance at preassigned thresholds is required to support these claims. ©RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Whitman and Moseley in this issue.

Imaging Glioblastoma Metabolism by Using Hyperpolarized [1-13C]Pyruvate Demonstrates Heterogeneity in Lactate Labeling: A Proof of Principle Study

Purpose To evaluate glioblastoma (GBM) metabolism by using hyperpolarized carbon 13 (13C) MRI to monitor the exchange of the hyperpolarized 13C label between injected [1-13C]pyruvate and tumor lactate and bicarbonate. Materials and Methods In this prospective study, seven treatment-naive patients (age [mean ± SD], 60 years ± 11; five men) with GBM were imaged at 3 T by using a dual-tuned 13C-hydrogen 1 head coil. Hyperpolarized [1-13C]pyruvate was injected, and signal was acquired by using a dynamic MRI spiral sequence. Metabolism was assessed within the tumor, in the normal-appearing brain parenchyma (NABP), and in healthy volunteers by using paired or unpaired t tests and a Wilcoxon signed rank test. The Spearman ρ correlation coefficient was used to correlate metabolite labeling with lactate dehydrogenase A (LDH-A) expression and some immunohistochemical markers. The Benjamini-Hochberg procedure was used to correct for multiple comparisons. Results The bicarbonate-to-pyruvate (BP) ratio was lower in the tumor than in the contralateral NABP (P < .01). The tumor lactate-to-pyruvate (LP) ratio was not different from that in the NABP (P = .38). The LP and BP ratios in the NABP were higher than those observed previously in healthy volunteers (P < .05). Tumor lactate and bicarbonate signal intensities were strongly correlated with the pyruvate signal intensity (ρ = 0.92, P < .001, and ρ = 0.66, P < .001, respectively), and the LP ratio was weakly correlated with LDH-A expression in biopsy samples (ρ = 0.43, P = .04). Conclusion Hyperpolarized 13C MRI demonstrated variation in lactate labeling in GBM, both within and between tumors. In contrast, bicarbonate labeling was consistently lower in tumors than in the surrounding NABP. Keywords: Hyperpolarized 13C MRI, Glioblastoma, Metabolism, Cancer, MRI, Neuro-oncology Supplemental material is available for this article. Published under a CC BY 4.0 license.