Evaluation of Combined Artificial Intelligence and Radiologist Assessment to Interpret Screening Mammograms.

Importance: Mammography screening currently relies on subjective human interpretation. Artificial intelligence (AI) advances could be used to increase mammography screening accuracy by reducing missed cancers and false positives. Objective: To evaluate whether AI can overcome human mammography interpretation limitations with a rigorous, unbiased evaluation of machine learning algorithms. Design, Setting, and Participants: In this diagnostic accuracy study conducted between September 2016 and November 2017, an international, crowdsourced challenge was hosted to foster AI algorithm development focused on interpreting screening mammography. More than 1100 participants comprising 126 teams from 44 countries participated. Analysis began November 18, 2016. Main Outcomes and Measurements: Algorithms used images alone (challenge 1) or combined images, previous examinations (if available), and clinical and demographic risk factor data (challenge 2) and output a score that translated to cancer yes/no within 12 months. Algorithm accuracy for breast cancer detection was evaluated using area under the curve and algorithm specificity compared with radiologists' specificity with radiologists' sensitivity set at 85.9% (United States) and 83.9% (Sweden). An ensemble method aggregating top-performing AI algorithms and radiologists' recall assessment was developed and evaluated. Results: Overall, 144 231 screening mammograms from 85 580 US women (952 cancer positive ≤12 months from screening) were used for algorithm training and validation. A second independent validation cohort included 166 578 examinations from 68 008 Swedish women (780 cancer positive). The top-performing algorithm achieved an area under the curve of 0.858 (United States) and 0.903 (Sweden) and 66.2% (United States) and 81.2% (Sweden) specificity at the radiologists' sensitivity, lower than community-practice radiologists' specificity of 90.5% (United States) and 98.5% (Sweden). Combining top-performing algorithms and US radiologist assessments resulted in a higher area under the curve of 0.942 and achieved a significantly improved specificity (92.0%) at the same sensitivity. Conclusions and Relevance: While no single AI algorithm outperformed radiologists, an ensemble of AI algorithms combined with radiologist assessment in a single-reader screening environment improved overall accuracy. This study underscores the potential of using machine learning methods for enhancing mammography screening interpretation.

Prognostic Significance of Cavitary Lung Nodules in Granulomatosis With Polyangiitis (Wegener's): A Clinical Imaging Study of 225 Patients.

OBJECTIVE: Granulomatosis with polyangiitis (Wegener's) (GPA) is a systemic necrotizing vasculitis in which pulmonary nodules are a common manifestation. Our study examined whether pulmonary nodules, and nodule type (solid versus cavitary), are associated with different disease manifestations and outcomes. METHODS: Demographic, clinical, biologic, and radiologic data at diagnosis and during followup and treatments of GPA patients followed at the Mount Sinai Hospital (Canada) Vasculitis Clinic were analyzed. Patients were separated by the absence of lung nodules, presence of solid nodules only, and presence of cavitary nodules (+/- solid nodules). The study outcomes included followup lung imaging, relapses, and deaths. RESULTS: Of 225 patients with GPA, 46 had solid nodules only and 44 had cavitary nodules at diagnosis. Demographic and clinical manifestations were similar in the patient subgroups at diagnosis. Cyclophosphamide (CYC) was used for induction after diagnosis in 76.7% of patients with cavitary nodules, compared with 64.7% of patients without nodules and 51.1% of patients with solid nodules (P = 0.04). The mean ± SD followup after diagnosis was 106.6 ± 92.6 months, and 6 of the patients died. In multivariable analysis, diagnosis before 2000 or pulmonary nodule cavitation at diagnosis were associated with relapse, with a hazard ratio of 0.38 (95% confidence interval [95% CI] 0.22-0.65; P < 0.001) and 1.53 (95% CI 1.00-2.33; P = 0.05), respectively, after adjustment for CYC use. CONCLUSION: The presence of cavitary nodules led to increased use of CYC but had no impact on survival. Relapse occurred more often, however, in patients with cavitary nodules than in those with solid nodules or no nodules, and should be studied in other cohorts.

Disturbed cartilage and joint homeostasis resulting from a loss of mitogen-inducible gene 6 in a mouse model of joint dysfunction.

OBJECTIVE: Mitogen-inducible gene 6 (MIG-6) regulates epidermal growth factor receptor (EGFR) signaling in synovial joint tissues. Whole-body knockout of the Mig6 gene in mice has been shown to induce osteoarthritis and joint degeneration. To evaluate the role of chondrocytes in this process, Mig6 was conditionally deleted from Col2a1-expressing cell types in the cartilage of mice. METHODS: Bone and cartilage in the synovial joints of cartilage-specific Mig6-deleted (knockout [KO]) mice and control littermates were compared. Histologic staining and immunohistochemical analyses were used to evaluate joint pathology as well as the expression of key extracellular matrix and regulatory proteins. Calcified tissue in synovial joints was assessed by micro-computed tomography (micro-CT) and whole-skeleton staining. RESULTS: Formation of long bones was found to be normal in KO animals. Cartilage thickness and proteoglycan staining of articular cartilage in the knee joints of 12-week-old KO mice were increased as compared to controls, with higher cellularity throughout the tissue. Radiopaque chondro-osseous nodules appeared in the knees of KO animals by 12 weeks of age and progressed to calcified bone-like tissue by 36 weeks of age. Nodules were also observed in the spine of 36-week-old animals. Erosion of bone at ligament entheses was evident by 12 weeks of age, by both histologic and micro-CT assessment. CONCLUSION: MIG-6 expression in chondrocytes is important for the maintenance of cartilage and joint homeostasis. Dysregulation of EGFR signaling in chondrocytes results in anabolic activity in cartilage, but erosion of ligament entheses and the formation of ectopic chondro-osseous nodules severely disturb joint physiology.

Targeted loss of the ATR-X syndrome protein in the limb mesenchyme of mice causes brachydactyly.

ATR-X syndrome is a rare genetic disorder caused by mutations in the ATRX gene. Affected individuals are cognitively impaired and display a variety of developmental abnormalities, including skeletal deformities. To investigate the function of ATRX during skeletal development, we selectively deleted the gene in the developing forelimb mesenchyme of mice. The absence of ATRX in the limb mesenchyme resulted in shorter digits, or brachydactyly, a defect also observed in a subset of ATR-X patients. This phenotype persisted until adulthood, causing reduced grip strength and altered gait in mutant mice. Examination of the embryonic ATRX-null forelimbs revealed a significant increase in apoptotic cell death, which could explain the reduced digit length. In addition, staining for the DNA damage markers γ-histone 2A family member X (γ-H2AX) and 53BP1 demonstrated a significant increase in the number of cells with DNA damage in the embryonic ATRX-null forepaw. Strikingly, only one large bright DNA damage event was observed per nucleus in proliferating cells. These large γ-H2AX foci were located in close proximity to the nuclear lamina and remained largely unresolved after cell differentiation. In addition, ATRX-depleted forelimb mesenchymal cells did not exhibit hypersensitivity to DNA fork-stalling compounds, suggesting that the nature as well as the response to DNA damage incurred by loss of ATRX in the developing limb fundamentally differs from other tissues. Our data suggest that DNA damage-induced apoptosis is a novel cellular mechanism underlying brachydactyly that might be relevant to additional skeletal syndromes.

Loss of ATRX does not confer susceptibility to osteoarthritis.

The chromatin remodelling protein ATRX is associated with the rare genetic disorder ATR-X syndrome. This syndrome includes developmental delay, cognitive impairment, and a variety of skeletal deformities. ATRX plays a role in several basic chromatin-mediated cellular events including DNA replication, telomere stability, gene transcription, and chromosome congression and cohesion during cell division. We have used a loss-of-function approach to directly investigate the role of Atrx in the adult skeleton in three different models of selective Atrx loss. We specifically targeted deletion of Atrx to the forelimb mesenchyme, to cartilage and to bone-forming osteoblasts. We previously demonstrated that loss of ATRX in forelimb mesenchyme causes brachydactyly while deletion in chondrocytes had minimal effects during development. We now show that targeted deletion of Atrx in osteoblasts causes minor dwarfism but does not recapitulate most of the skeletal phenotypes seen in ATR-X syndrome patients. In adult mice from all three models, we find that joints lacking Atrx are not more susceptible to osteoarthritis, as determined by OARSI scoring and immunohistochemistry. These results indicate that while ATRX plays limited roles during early stages of skeletal development, deficiency of the protein in adult tissues does not confer susceptibility to osteoarthritis.