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.

The prevalence of femoroacetabular impingement anatomy in Division 1 aquatic athletes who tread water.

Femoroacetabular impingement (FAI) is a disorder that causes hip pain and disability in young patients, particularly athletes. Increased stress on the hip during development has been associated with increased risk of cam morphology. The specific forces involved are unclear, but may be due to continued rotational motion, like the eggbeater kick. The goal of this prospective cohort study was to use magnetic resonance imaging (MRI) to identify the prevalence of FAI anatomy in athletes who tread water and compare it to the literature on other sports. With university IRB approval, 20 Division 1 water polo players and synchronized swimmers (15 female, 5 male), ages 18-23 years (mean age 20.7 ± 1.4), completed the 33-item International Hip Outcome Tool and underwent non-contrast MRI scans of both hips using a 3 Tesla scanner. Recruitment was based on sport, with both symptomatic and asymptomatic individuals included. Cam and pincer morphology were identified. The Wilcoxon Signed-Rank/Rank Sum tests were used to assess outcomes. Seventy per cent (14/20) of subjects reported pain in their hips yet only 15% (3/20) sought clinical evaluation. Cam morphology was present in 67.5% (27/40) of hips, while 22.5% (9/40) demonstrated pincer morphology. The prevalence of cam morphology in water polo players and synchronized swimmers is greater than that reported for the general population and at a similar level as some other sports. From a clinical perspective, acknowledgment of the high prevalence of cam morphology in water polo players and synchronized swimmers should be considered when these athletes present with hip pain.

Risk factors and the natural history of accelerated knee osteoarthritis: a narrative review.

BACKGROUND: Osteoarthritis is generally a slowly progressive disorder. However, at least 1 in 7 people with incident knee osteoarthritis develop an abrupt progression to advanced-stage radiographic disease, many within 12 months. We summarize what is known - primarily based on findings from the Osteoarthritis Initiative - about the risk factors and natural history of accelerated knee osteoarthritis (AKOA) - defined as a transition from no radiographic knee osteoarthritis to advanced-stage disease < 4 years - and put these findings in context with typical osteoarthritis (slowly progressing disease), aging, prior case reports/series, and relevant animal models. Risk factors in the 2 to 4 years before radiographic manifestation of AKOA (onset) include older age, higher body mass index, altered joint alignment, contralateral osteoarthritis, greater pre-radiographic disease burden (structural, symptoms, and function), or low fasting glucose. One to 2 years before AKOA onset people often exhibit rapid articular cartilage loss, larger bone marrow lesions and effusion-synovitis, more meniscal pathology, slower chair-stand or walking pace, and increased global impact of arthritis than adults with typical knee osteoarthritis. Increased joint symptoms predispose a person to new joint trauma, which for someone who develops AKOA is often characterized by a destabilizing meniscal tear (e.g., radial or root tear). One in 7 people with AKOA onset subsequently receive a knee replacement during a 9-year period. The median time from any increase in radiographic severity to knee replacement is only 2.3 years. Despite some similarities, AKOA is different than other rapidly progressive arthropathies and collapsing these phenomena together or extracting results from one type of osteoarthritis to another should be avoided until further research comparing these types of osteoarthritis is conducted. Animal models that induce meniscal damage in the presence of other risk factors or create an incongruent distribution of loading on joints create an accelerated form of osteoarthritis compared to other models and may offer insights into AKOA. CONCLUSION: Accelerated knee osteoarthritis is unique from typical knee osteoarthritis. The incidence of AKOA in the Osteoarthritis Initiative and Chingford Study is substantial. AKOA needs to be taken into account and studied in epidemiologic studies and clinical trials.

Dynamic contrast-enhanced computed tomography for the diagnosis of solitary pulmonary nodules: a systematic review and meta-analysis.

INTRODUCTION: A systematic review and meta-analysis were performed to determine the diagnostic performance of dynamic contrast-enhanced computed tomography (DCE-CT) for the differentiation between malignant and benign pulmonary nodules. METHODS: Ovid MEDLINE and EMBASE were searched for studies published up to October 2018 on the diagnostic accuracy of DCE-CT for the characterisation of pulmonary nodules. For the index test, studies with a minimum of a pre- and post-contrast computed tomography scan were evaluated. Studies with a reference standard of biopsy for malignancy, and biopsy or 2-year follow-up for benign disease were included. Study bias was assessed using QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies). The sensitivities, specificities, and diagnostic odds ratios were determined along with 95% confidence intervals (CIs) using a bivariate random effects model. RESULTS: Twenty-three studies were included, including 2397 study participants with 2514 nodules of which 55.3% were malignant (1389/2514). The pooled accuracy results were sensitivity 94.8% (95% CI 91.5; 96.9), specificity 75.5% (69.4; 80.6), and diagnostic odds ratio 56.6 (24.2-88.9). QUADAS 2 assessment showed intermediate/high risk of bias in a large proportion of the studies (52-78% across the domains). No difference was present in sensitivity or specificity between subgroups when studies were split based on CT technique, sample size, nodule size, or publication date. CONCLUSION: DCE-CT has a high diagnostic accuracy for the diagnosis of pulmonary nodules although study quality was indeterminate in a large number of cases. KEY POINTS: • The pooled accuracy results were sensitivity 95.1% and specificity 73.8% although individual studies showed wide ranges of values. • This is comparable to the results of previous meta-analyses of PET/CT (positron emission tomography/computed tomography) diagnostic accuracy for the diagnosis of solitary pulmonary nodules. • Robust direct comparative accuracy and cost-effectiveness studies are warranted to determine the optimal use of DCE-CT and PET/CT in the diagnosis of SPNs.

Accelerated Knee Osteoarthritis Is Characterized by Destabilizing Meniscal Tears and Preradiographic Structural Disease Burden.

OBJECTIVE: To determine whether accelerated knee osteoarthritis (KOA) is preceded by, and characterized over time by, destabilizing meniscal tears or other pathologic changes. METHODS: We selected 3 sex-matched groups of subjects from the first 48 months of the Osteoarthritis Initiative, comprising adults who had a knee without KOA (Kellgren/Lawrence [K/L] radiographic grade <2) at baseline. Subjects in the accelerated KOA group developed KOA of K/L grade ≥3, those with typical KOA showed increased K/L radiographic scores, and those with no KOA had the same K/L grade over time. An index visit was the visit when the radiographic criteria for accelerated KOA and typical KOA were met (the no KOA group was matched to the accelerated KOA group). The observation period was up to 2 years before and after an index visit. Radiologists reviewed magnetic resonance (MR) images of the index knee and identified destabilizing meniscal tears (root tears, radial tears, complex tears), miscellaneous pathologic features (acute ligamentous or tendinous injuries, attrition, subchondral insufficiency fractures, other incidental findings), and meniscal damage in >2 of 6 regions (3 regions per meniscus: anterior horn, body, posterior horn). In addition, bone marrow lesions (BMLs) and cartilage damage on MR images were quantified. Linear mixed regression models were performed to analyze the results. RESULTS: At 1 year before the index visit, >75% of adults with accelerated KOA had meniscal damage in ≥2 regions (odds ratio 3.19 [95% confidence interval 1.70-5.97] versus adults with typical KOA). By the index visit, meniscal damage in ≥2 regions was ubiquitous in adults with accelerated KOA, including 42% of subjects having evidence of a destabilizing meniscal tear (versus 14% of subjects with typical KOA). These changes corresponded to findings of larger BMLs and greater cartilage loss in the accelerated KOA group. CONCLUSION: Accelerated KOA is characterized by destabilizing meniscal tears in a knee compromised by meniscal damage in >2 regions, and also characterized by the presence of large BMLs and greater cartilage loss.