Regional european genetic ancestry predicts type I interferon level and risk of severe viral infection.

BACKGROUND: Viral infection outcomes vary widely between individuals, ranging from mild symptoms to severe organ failure and death, and it is clear that host genetic factors play a role in this variability. Type I interferon (IFN) is a critical anti-viral cytokine, and we have previously noted differences in type I IFN levels between world populations. METHODS: In this study, we investigate the interrelationship between regional European genetic ancestry, type I IFN levels, and severe viral infection outcomes. RESULTS: In cohorts of European ancestry lupus patients living in Europe, we noted higher IFN in the Northwestern populations as compared to Southeastern populations. In an independent cohort of European ancestry lupus patients from the United States with varying proportional regional European genetic admixture, we observed the same Northwest vs. Southeast European ancestry IFN gradient. We developed a model to predict type I IFN level based on regional European ancestry (AUC = 0.73, p = 6.1e-6). Examining large databases containing serious viral outcomes data, we found that lower predicted IFN in the corresponding European country was significantly correlated with increased viral infection fatality rate, including COVID-19, viral hepatitis, and HIV [Correlation coefficients: -0.79 (p = 4e-2), -0.94 (p = 6e-3), and -0.96 (p = 8e-2) respectively]. CONCLUSIONS: This association between predicted type I IFN level and viral outcome severity suggests a potential causal relationship, as greater intrinsic type I IFN is beneficial in host defense against viruses. Genetic testing could provide insight into individual and population level risk of fatality due to viruses prior to infection, across a wide range of viral pathogens.

Validating the Fracture Risk Assessment Tool Score in a US Population-Based Study of Patients With Rheumatoid Arthritis.

OBJECTIVE: The World Health Organization fracture risk assessment tool (FRAX) algorithm for risk prediction of major osteoporotic and hip fractures accounts for several risk factors, including rheumatoid arthritis (RA), since individuals with RA have an excess burden of fractures. FRAX has not been validated in population-based RA cohorts in the US. We aimed to determine the accuracy of FRAX predictions for individuals with RA in the US. METHODS: This retrospective population-based cohort study included residents of Olmsted County, Minnesota, who were followed until death, migration, or last medical record review. Each patient with RA (1987 American College of Rheumatology criteria met in 1980-2007, age 40-89 years) was matched 1:1 on age and sex to an individual without RA from the same underlying population. Ten-year predictions for major osteoporotic and hip fractures were estimated using the FRAX tool. Fractures were ascertained through follow-up, truncated at 10 years. Standardized incidence ratios (SIRs) and 95% CI were calculated to compare observed and predicted fractures. RESULTS: The study included 662 patients with RA and 658 non-RA comparators (66.8% vs 66.9% female and a mean age of 60.6 vs 60.5 years, respectively). Among patients with RA, 76 major osteoporotic fractures and 21 hip fractures were observed during follow-up (median follow-up: 9.0 years) compared to 67.0 predicted major osteoporotic fractures (SIR 1.13, 95% CI 0.91-1.42) and 23.3 predicted hip fractures (SIR 0.90, 95% CI 0.59-1.38). The observed and predicted major osteoporotic and hip fracture risks were similar for patients with RA and non-RA comparators. CONCLUSION: The FRAX tool is an accurate method for estimating major osteoporotic and hip fracture risk in patients with RA.

ST-V-Net: incorporating shape prior into convolutional neural networks for proximal femur segmentation.

We aim to develop a deep-learning-based method for automatic proximal femur segmentation in quantitative computed tomography (QCT) images. We proposed a spatial transformation V-Net (ST-V-Net), which contains a V-Net and a spatial transform network (STN) to extract the proximal femur from QCT images. The STN incorporates a shape prior into the segmentation network as a constraint and guidance for model training, which improves model performance and accelerates model convergence. Meanwhile, a multi-stage training strategy is adopted to fine-tune the weights of the ST-V-Net. We performed experiments using a QCT dataset which included 397 QCT subjects. During the experiments for the entire cohort and then for male and female subjects separately, 90% of the subjects were used in ten-fold stratified cross-validation for training and the rest of the subjects were used to evaluate the performance of models. In the entire cohort, the proposed model achieved a Dice similarity coefficient (DSC) of 0.9888, a sensitivity of 0.9966 and a specificity of 0.9988. Compared with V-Net, the Hausdorff distance was reduced from 9.144 to 5.917 mm, and the average surface distance was reduced from 0.012 to 0.009 mm using the proposed ST-V-Net. Quantitative evaluation demonstrated excellent performance of the proposed ST-V-Net for automatic proximal femur segmentation in QCT images. In addition, the proposed ST-V-Net sheds light on incorporating shape prior to segmentation to further improve the model performance.

Toto, we're not in Kansas anymore: First reported case of M. persicum septic arthritis.

In this report, we describe a case of septic arthritis caused by the newly described Mycobacterium persicum (formerly Mycobacterium kansasii complex). The patient's only significant exposure was home gardening. To our knowledge, this represents the first documented case of M. persicum infection in the United States and first septic arthritis.

Native joint polyarticular septic arthritis secondary to disseminated Ureaplasma urealyticum infection in a patient on rituximab therapy with hypogammaglobulinemia: A Case Report.

Ureaplasma urealyticum and Ureaplasma parvum are important causes of septic arthritis in patients with hypogammaglobulinemia. The diagnosis can be challenging, leading to prolonged illness and increased morbidity, and mortality. This is driven by the complex growth media requirements of Ureaplasma species and the difficulty in identifying the organisms on routine culture media. Herein, we present a case of native joint polyarticular septic arthritis and vertebral infection secondary to disseminated U. urealyticum in a patient maintained on rituximab. The diagnosis was established through a positive species-specific U. urealyticum polymerase chain reaction (PCR) after a meticulous workup including synovial fluid biopsy, cultures and broad-range bacterial PCR returned negative. Septic arthritis caused by Ureaplasma species should be considered in the differential diagnosis especially in immunocompromised patients with hypogammaglobulinemia, even if the initial microbiological workup is non-revealing. Delayed diagnosis and treatment are associated with increased morbidity.

mTORC2 contributes to systemic autoimmunity.

The development of many systemic autoimmune diseases, including systemic lupus erythematosus, is associated with overactivation of the type I interferon (IFN) pathway, lymphopenia and increased follicular helper T (Tfh)-cell differentiation. However, the cellular and molecular mechanisms underlying these immunological perturbations remain incompletely understood. Here, we show that the mechanistic target of rapamycin complex 2 (mTORC2) promotes Tfh differentiation and disrupts Treg homeostasis. Inactivation of mTORC2 in total T cells, but not in Tregs, greatly ameliorated the immunopathology in a systemic autoimmunity mouse model. This was associated with reduced Tfh differentiation, B-cell activation, and reduced T-cell glucose metabolism. Finally, we show that type I IFN can synergize with TCR ligation to activate mTORC2 in T cells, which partially contributes to T-cell lymphopenia. These data indicate that mTORC2 may act as downstream of type I IFN, TCR and costimulatory receptor ICOS, to promote glucose metabolism, Tfh differentiation, and T-cell lymphopenia, but not to suppress Treg function in systemic autoimmunity. Our results suggest that mTORC2 might be a rational target for systemic autoimmunity treatment.

Rituximab for the Treatment of Refractory Cardiac Sarcoidosis: A Single-Center Experience.

BACKGROUND: We sought to examine the effect of anti-B-cell therapy (rituximab) on cardiac inflammation and function in corticosteroid-refractory cardiac sarcoidosis. Cardiac sarcoidosis (CS) is a rare cause of cardiomyopathy characterized by granulomatous inflammation involving the myocardium. Although typically responsive to corticosteroid treatment, there is a critical need for identifying effective steroid-sparing agents for disease control. Despite increasing evidence on the role of B cells in the pathogenesis of sarcoidosis, there is limited data on the efficacy of anti-B-cell therapy, specifically rituximab, for controlling CS. METHODS AND RESULTS: We reviewed the clinical experience at a tertiary care referral center of all patients with CS who received rituximab after failing to improve with initial immunosuppression therapy, which included corticosteroids. Fluorodeoxyglucose positron emission tomography (FDG PET/CT) images before and after rituximab treatment were evaluated. All images were interpreted by 2 experienced nuclear medicine trained physicians. We identified 7 patients (5 men, 2 women; mean age at diagnosis, 49.0 ± 7.9 years) with active CS who were treated with rituximab. The median length of follow-up was 5.1 years. All individuals, but 1, had received prior steroid-sparing agents in addition to corticosteroids. Rituximab was administered either as 1000 mg intravenously ×1 or ×2 doses, separated by 2 weeks. Repeat dosing, if appropriate, was considered after 6 months. All tolerated the infusions well. Inflammation as assessed by maximum standardized uptake value on cardiac FDG PET/CT uptake significantly decreased in 6 of 7 patients (median 6.0-4.5, Wilcoxon signed rank z -1.8593, W 3), whereas the left ventricular ejection fraction improved or stabilized in 4 patients but decreased in 3. The mean left ventricular ejection fraction was 40.1% and 43.3% before and after treatment, respectively (P = .28). Three patients reported improved physical capacity, and 5 patients showed improved arrhythmic burden on Holter monitoring or implantable cardioverter-defibrillator interrogation. One patient subsequently developed a fungal catheter-associated infection and sepsis requiring discontinuation. CONCLUSIONS: Rituximab was well-tolerated and seemed to decrease inflammation, as assessed by cardiac FDG PET/CT in all but 1 patient with active CS. These data suggest that rituximab may be a promising therapeutic option for CS, which deserves merits further study.

High Systemic Type I Interferon Activity Is Associated With Active Class III/IV Lupus Nephritis.

OBJECTIVE: Previous studies suggest a link between high serum type I interferon (IFN) and lupus nephritis (LN). We determined whether serum IFN activity is associated with subtypes of LN and studied renal tissues and cells to understand the effect of IFN in LN. METHODS: Two hundred and twenty-one patients with systemic lupus erythematosus were studied. Serum IFN activity was measured by WISH bioassay. mRNA in situ hybridization was used in renal tissue to measure expression of the representative IFN-induced gene, IFN-induced protein with tetratricopeptide repeats-1 (IFIT1), and the plasmacytoid dendritic cell (pDC) marker gene C-type lectin domain family-4 member C (CLEC4C). Podocyte cell line gene expression was measured by real-time PCR. RESULTS: Class III/IV LN prevalence was significantly increased in patients with high serum IFN compared with those with low IFN (odds ratio 5.40, P = 0.009). In multivariate regression models, type I IFN was a stronger predictor of class III/IV LN than complement C3 or anti-dsDNA antibody, and could account for the association of these variables with LN. IFIT1 expression was increased in all classes of LN, but most in the glomerular areas of active class III/IV LN kidneys. IFIT1 expression was not closely colocalized with pDCs. IFN directly activated podocyte cell lines to induce chemokines and proapoptotic molecules. CONCLUSION: Systemic high IFN is involved in the pathogenesis of severe LN. We did not find colocalization of pDCs with IFN signature in renal tissue, and instead observed the greatest intensity of the IFN signature in glomerular areas, which could suggest a blood source of IFN.

Single-cell expression quantitative trait loci (eQTL) analysis of SLE-risk loci in lupus patient monocytes.

BACKGROUND: We performed expression quantitative trait locus (eQTL) analysis in single classical (CL) and non-classical (NCL) monocytes from patients with systemic lupus erythematosus (SLE) to quantify the impact of well-established genetic risk alleles on transcription at single-cell resolution. METHODS: Single-cell gene expression was quantified using qPCR in purified monocyte subpopulations (CD14++CD16- CL and CD14dimCD16+ NCL) from SLE patients. Novel analysis methods were used to control for the within-person correlations observed, and eQTLs were compared between cell types and risk alleles. RESULTS: The SLE-risk alleles demonstrated significantly more eQTLs in NCLs as compared to CLs (p = 0.0004). There were 18 eQTLs exclusive to NCL cells, 5 eQTLs exclusive to CL cells, and only one shared eQTL, supporting large differences in the impact of the risk alleles between these monocyte subsets. The SPP1 and TNFAIP3 loci were associated with the greatest number of transcripts. Patterns of shared influence in which different SNPs impacted the same transcript also differed between monocyte subsets, with greater evidence for synergy in NCL cells. IRF1 expression demonstrated an on/off pattern, in which expression was zero in all of the monocytes studied from some individuals, and this pattern was associated with a number of SLE risk alleles. We observed corroborating evidence of this IRF1 expression pattern in public data sets. CONCLUSIONS: We document multiple SLE-risk allele eQTLs in single monocytes which differ greatly between CL and NCL subsets. These data support the importance of the SPP1 and TNFAIP3 risk variants and the IRF1 transcript in SLE patient monocyte function.

Determinants of Bone Material Strength and Cortical Porosity in Patients with Type 2 Diabetes Mellitus.

CONTEXT: Reduced bone material strength index (BMSi) and increased cortical porosity (CtPo) have emerged as potentially contributing to fracture risk in type 2 diabetes mellitus (T2DM) patients. OBJECTIVE: To determine whether BMSi or CtPo are related to other diabetic complications. DESIGN: Cross-sectional observational study. SETTING: Subjects recruited from a random sample of southeast Minnesota residents. PARTICIPANTS: A total of 171 T2DM patients (mean age, 68.8 years) and 108 age-matched nondiabetic controls (mean age, 67.3 years). MAIN MEASURES: Bone material strength index was measured using microindentation, skin advanced glycation end-products (AGEs) measured using autofluorescence, high-resolution peripheral quantitative computed tomography at the distal radius and tibia, assessment of diabetic microvascular complications including urine microalbuminuria, retinopathy, neuropathy, and vascular disease (ankle brachial index and transcutaneous oxygen tension [TcPO2]). All analyses were adjusted for age, sex, and body mass index. RESULTS: Skin AGEs were negatively correlated with the BMSi in both T2DM (r = -0.30, P < 0.001) and control (r = -0.23, P = 0.020) subjects. In relating diabetic complications to CtPo, we found that T2DM patients with clinically significant peripheral vascular disease (TcPO2 ≤ 40 mm Hg) had higher (+21.0%, P = 0.031) CtPo at the distal tibia as compared to controls; in these subjects, CtPo was negatively correlated with TcPO2 at both the distal tibia (r = -0.39, P = 0.041) and radius (r = -0.41, P = 0.029). CONCLUSIONS: Our findings demonstrate that bone material properties are related to AGE accumulation regardless of diabetes status, while CtPo in T2DM patients is linked to TcPO2, a measure of microvascular blood flow.

Posterior single-stepping thresholds are prospectively related to falls in older women.

BACKGROUND: Falls are a leading cause of injury in older women. Stepping thresholds quantify balance-reaction capabilities. It is unclear how such evaluations predict falls in comparison to, or as a complement to, other objective measures of gait, standing postural control, strength, and balance confidence. AIMS: The objective of this study was to determine if stepping thresholds are prospectively related to falls in older women. METHODS: For this prospective cohort study, 125 ambulatory, community-dwelling women, age ≥ 65 years were recruited. Using a treadmill to deliver perturbations to standing participants, we determined anteroposterior single- and multiple-stepping thresholds. Here, thresholds represent the minimum perturbation magnitudes that consistently evoke one step or multiple steps. In addition, gait kinematics, obstacle-crossing kinematics, standing sway measures, unipedal stance time, the functional reach, lower extremity isometric strength, grip strength, balance confidence, and fall history were evaluated. Falls were prospectively recorded for one year. RESULTS: Seventy-four participants (59%) fell at least once. Posterior single-stepping thresholds were the only outcome that predicted future fall status (OR = 1.50, 95% CI 1.01-2.28; AUC = .62). A multivariate approach added postural sway with eyes closed as a second predictive variable, although predictive abilities were not meaningfully improved. DISCUSSION: These results align with the previous evidence that reactive balance is a prospective indicator of fall risk. Unlike previous studies, strength scaled to body size did not contribute to fall prediction. CONCLUSION: Posterior single-stepping thresholds held a significant relationship with future fall status. This relationship was independent of, and superior to that of, other measures of standing balance, gait, strength, and balance confidence.

Unsupervised machine learning for the discovery of latent disease clusters and patient subgroups using electronic health records.

Machine learning has become ubiquitous and a key technology on mining electronic health records (EHRs) for facilitating clinical research and practice. Unsupervised machine learning, as opposed to supervised learning, has shown promise in identifying novel patterns and relations from EHRs without using human created labels. In this paper, we investigate the application of unsupervised machine learning models in discovering latent disease clusters and patient subgroups based on EHRs. We utilized Latent Dirichlet Allocation (LDA), a generative probabilistic model, and proposed a novel model named Poisson Dirichlet Model (PDM), which extends the LDA approach using a Poisson distribution to model patients' disease diagnoses and to alleviate age and sex factors by considering both observed and expected observations. In the empirical experiments, we evaluated LDA and PDM on three patient cohorts, namely Osteoporosis, Delirium/Dementia, and Chronic Obstructive Pulmonary Disease (COPD)/Bronchiectasis Cohorts, with their EHR data retrieved from the Rochester Epidemiology Project (REP) medical records linkage system, for the discovery of latent disease clusters and patient subgroups. We compared the effectiveness of LDA and PDM in identifying disease clusters through the visualization of disease representations. We tested the performance of LDA and PDM in differentiating patient subgroups through survival analysis, as well as statistical analysis of demographics and Elixhauser Comorbidity Index (ECI) scores in those subgroups. The experimental results show that the proposed PDM could effectively identify distinguished disease clusters based on the latent patterns hidden in the EHR data by alleviating the impact of age and sex, and that LDA could stratify patients into differentiable subgroups with larger p-values than PDM. However, those subgroups identified by LDA are highly associated with patients' age and sex. The subgroups discovered by PDM might imply the underlying patterns of diseases of greater interest in epidemiology research due to the alleviation of age and sex. Both unsupervised machine learning approaches could be leveraged to discover patient subgroups using EHRs but with different foci.

Interferon Chemokine Score and Other Cytokine Measures Track With Changes in Disease Activity in Patients With Juvenile and Adult Dermatomyositis.

OBJECTIVE: Our aim was to identify cytokines and chemokines in patients with adult dermatomyositis (DM) and juvenile dermatomyositis (JDM) that predict changes in disease activity. METHODS: Multiplexed immunoassays (Meso Scale Discovery) enabled simultaneous measurement of interferon (IFN)-regulated chemokines and other pro- and anti-inflammatory cytokines specific to differentiation of specific T-cell and innate pathways. Cytokine scores were computed for IFNCK (IP-10, MCP-1), Th1 (IFNÉ£, TNFα, and IL2), Th2 (IL4, IL10, IL12, and IL 13), Th17 (IL6, IL17, IL1ß), macrophage (MIP-1α, MIP-1ß, IL8), and regulatory (IL10, TNFα) factors. Spearman correlation and mixed models were used to examine whether cytokines at a previous visit predict change in disease activity at the next visit. RESULTS: The study included 36 patients (16 DM and 20 JDM) with at least two visits (87 patient intervals between two visits). Mean age (SD) at inclusion was 56.9 (18.4) years for DM and 10.8 (6.6) years in JDM, 67% of patients were female, 89% Caucasian. The mean (SD) physician global, muscle and extra-muscular disease activity Visual Analog Scale scores at inclusion were 41 (26), 36 (30), and 34 (21) mm, respectively. The change in IFN score from one visit to the next was associated with the change in physician global (P = 0.010) and extramuscular (P < 0.001) disease activity scores. Preliminary results revealed significant correlations of previous IFNCK score and IL-6 with subsequent disease activity measures, but after adjustment for multiple visits per patient, these associations did not reach statistical significance. CONCLUSION: There is a potential relationship between IFNCK and other cytokine scores seen in adult and juvenile DM with future disease states.

Type I Interferon Predicts an Alternate Immune System Phenotype in Systemic Lupus Erythematosus.

OBJECTIVE: Type I interferon (IFN) is important to systemic lupus erythematosus (SLE) pathogenesis, but it is not clear how chronic elevations in IFN alter immune function. We compared cytokine responses after whole blood stimulation with Toll-like receptor (TLR) agonists in high- and low-IFN SLE patient subgroups. METHODS: SLE patients and nonautoimmune controls were recruited, and SLE patients were categorized as either high or low IFN. Whole blood was dispensed into tubes coated with lipopolysaccharide (LPS), oligonucleotides with cytosine-guanine repeats, Resiquimod, IFN-α, and IFN-α + LPS. Cytokine production in patient sera and after whole blood TLR stimulation was measured by multiplex assay, and type I IFN was assessed using a functional assay. RESULTS: Circulating plasmacytoid dendritic cell numbers were specifically reduced in high-IFN SLE patients and not in low-IFN SLE patients. In serum, we observed that the correlations between cytokines in serum differed to a much greater degree between the high- and low-IFN groups (P < 0.0001) than the absolute cytokine levels differed between these same groups. In stimulated conditions, the high-IFN patients had less cytokine production in response to TLR ligation than the low-IFN SLE patients. LPS produced the most diverse response, and a number of interactions between type I IFN and LPS were observed. CONCLUSION: We find striking differences in resting and stimulated cytokine patterns in high- vs. low-IFN SLE patients, which supports the biological importance of these patient subsets. These data could inform personalized treatment approaches and the pathogenesis of SLE flare following infection.

Glial Fibrillary Acidic Protein (GFAP) Autoimmunity in the Setting of Seropositive Rheumatoid Arthritis Treated With Etanercept.

INTRODUCTION: Glial fibrillary acidic protein (GFAP) immunoglobulin G is a recently discovered biomarker of an autoimmune central nervous system disorder characterized by a steroid-responsive meningoencephalomyelitis. CASE REPORT: A 63-year-old man with rheumatoid arthritis on etanercept presented with steroid-responsive subacute encephalopathy and foot drop. Brain and sural nerve biopsies demonstrated a T-cell perivascular infiltrate. Cerebrospinal fluid studies 18 months into the course of the illness demonstrated a GFAP antibody on mouse tissue immunofluorescence confirmed by cell-based assay. The patient was treated with steroids and cyclophosphamide leading to resolution of his symptoms. CONCLUSION: This case expands on the previously reported cases of GFAP immunoglobulin G autoimmunity by describing an associated inflammatory large fiber peripheral neuropathy.

The Epidemiology of Antiphospholipid Syndrome: A Population-Based Study.

OBJECTIVE: To estimate the annual incidence and prevalence of and frequency of mortality associated with antiphospholipid syndrome (APS). METHODS: An inception cohort of patients with incident APS in 2000-2015 from a geographically well-defined population was identified based on comprehensive individual medical records review. All cases met the 2006 Sydney criteria for APS (primary definition) or had a diagnosis of APS confirmed by physician consensus (secondary definition). Levels of lupus anticoagulant, IgM and IgG anticardiolipin antibodies, and anti-ß2-glycoprotein I antibodies were tested in a centralized laboratory. Incidence rates were age- and sex-adjusted to the 2010 US white population. Prevalence estimates were obtained from the incidence rates, assuming that there was no increased mortality associated with APS and that migration in or out of the area was independent of disease status. RESULTS: Among this cohort in 2000-2015, 33 cases of incident APS, as defined by the Sydney criteria, were identified (mean age of patients 54.2 years; 55% female, 97% white). The annual incidence of APS in adults ages ≥18 years was 2.1 (95% confidence interval [95% CI] 1.4-2.8) per 100,000 population. Incidence rates were similar in both sexes. The estimated prevalence of APS was 50 (95% CI 42-58) per 100,000 population, and was similar in both sexes. Six patients (18%) had a concurrent diagnosis of systemic lupus erythematosus. The most frequent clinical manifestation was deep vein thrombosis. The overall frequency of mortality among patients with APS was not significantly different from that in the general population (standardized mortality ratio 1.61, 95% CI 0.74-3.05). CONCLUSION: APS occurred in ~2 persons per 100,000 population per year. The estimated prevalence was 50 per 100,000 population. Overall mortality was not notably different from that observed in the general population.

Natural language processing of radiology reports for identification of skeletal site-specific fractures.

BACKGROUND: Osteoporosis has become an important public health issue. Most of the population, particularly elderly people, are at some degree of risk of osteoporosis-related fractures. Accurate identification and surveillance of patient populations with fractures has a significant impact on reduction of cost of care by preventing future fractures and its corresponding complications. METHODS: In this study, we developed a rule-based natural language processing (NLP) algorithm for identification of twenty skeletal site-specific fractures from radiology reports. The rule-based NLP algorithm was based on regular expressions developed using MedTagger, an NLP tool of the Apache Unstructured Information Management Architecture (UIMA) pipeline to facilitate information extraction from clinical narratives. Radiology notes were retrieved from the Mayo Clinic electronic health records data warehouse. We developed rules for identifying each fracture type according to physicians' knowledge and experience, and refined these rules via verification with physicians. This study was approved by the institutional review board (IRB) for human subject research. RESULTS: We validated the NLP algorithm using the radiology reports of a community-based cohort at Mayo Clinic with the gold standard constructed by medical experts. The micro-averaged results of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1-score of the proposed NLP algorithm are 0.930, 1.0, 1.0, 0.941, 0.961, respectively. The F1-score is 1.0 for 8 fractures, and above 0.9 for a total of 17 out of 20 fractures (85%). CONCLUSIONS: The results verified the effectiveness of the proposed rule-based NLP algorithm in automatic identification of osteoporosis-related skeletal site-specific fractures from radiology reports. The NLP algorithm could be utilized to accurately identify the patients with fractures and those who are also at high risk of future fractures due to osteoporosis. Appropriate care interventions to those patients, not only the most at-risk patients but also those with emerging risk, would significantly reduce future fractures.

Hip load capacity cut-points for Astronaut Skeletal Health NASA Finite Element Strength Task Group Recommendations.

Concerns raised at a 2010 Bone Summit held for National Aeronautics and Space Administration Johnson Space Center led experts in finite element (FE) modeling for hip fracture prediction to propose including hip load capacity in the standards for astronaut skeletal health. The current standards for bone are based upon areal bone mineral density (aBMD) measurements by dual X-ray absorptiometry (DXA) and an adaptation of aBMD cut-points for fragility fractures. Task Group members recommended (i) a minimum permissible outcome limit (POL) for post-mission hip bone load capacity, (ii) use of FE hip load capacity to further screen applicants to astronaut corps, (iii) a minimum pre-flight standard for a second long-duration mission, and (iv) a method for assessing which post-mission physical activities might increase an astronaut's risk for fracture after return. QCT-FE models of eight astronaut were analyzed using nonlinear single-limb stance (NLS) and posterolateral fall (NLF) loading configurations. QCT data from the Age Gene/Environment Susceptibility (AGES) Reykjavik cohort and the Rochester Epidemiology Project were analyzed using identical modeling procedures. The 75th percentile of NLS hip load capacity for fractured elderly males of the AGES cohort (9537N) was selected as a post-mission POL. The NLF model, in combination with a Probabilistic Risk Assessment tool, was used to assess the likelihood of exceeding the hip load capacity during post-flight activities. There was no recommendation to replace the current DXA-based standards. However, FE estimation of hip load capacity appeared more meaningful for younger, physically active astronauts and was recommended to supplement aBMD cut-points.

A clinical text classification paradigm using weak supervision and deep representation.

BACKGROUND: Automatic clinical text classification is a natural language processing (NLP) technology that unlocks information embedded in clinical narratives. Machine learning approaches have been shown to be effective for clinical text classification tasks. However, a successful machine learning model usually requires extensive human efforts to create labeled training data and conduct feature engineering. In this study, we propose a clinical text classification paradigm using weak supervision and deep representation to reduce these human efforts. METHODS: We develop a rule-based NLP algorithm to automatically generate labels for the training data, and then use the pre-trained word embeddings as deep representation features for training machine learning models. Since machine learning is trained on labels generated by the automatic NLP algorithm, this training process is called weak supervision. We evaluat the paradigm effectiveness on two institutional case studies at Mayo Clinic: smoking status classification and proximal femur (hip) fracture classification, and one case study using a public dataset: the i2b2 2006 smoking status classification shared task. We test four widely used machine learning models, namely, Support Vector Machine (SVM), Random Forest (RF), Multilayer Perceptron Neural Networks (MLPNN), and Convolutional Neural Networks (CNN), using this paradigm. Precision, recall, and F1 score are used as metrics to evaluate performance. RESULTS: CNN achieves the best performance in both institutional tasks (F1 score: 0.92 for Mayo Clinic smoking status classification and 0.97 for fracture classification). We show that word embeddings significantly outperform tf-idf and topic modeling features in the paradigm, and that CNN captures additional patterns from the weak supervision compared to the rule-based NLP algorithms. We also observe two drawbacks of the proposed paradigm that CNN is more sensitive to the size of training data, and that the proposed paradigm might not be effective for complex multiclass classification tasks. CONCLUSION: The proposed clinical text classification paradigm could reduce human efforts of labeled training data creation and feature engineering for applying machine learning to clinical text classification by leveraging weak supervision and deep representation. The experimental experiments have validated the effectiveness of paradigm by two institutional and one shared clinical text classification tasks.