Identifying pediatric heart murmurs and distinguishing innocent from pathologic using deep learning.

OBJECTIVE: To develop a deep learning algorithm to perform multi-class classification of normal pediatric heart sounds, innocent murmurs, and pathologic murmurs. METHODS: We prospectively enrolled children under age 18 being evaluated by the Division of Pediatric Cardiology. Parents provided consent for a deidentified recording of their child's heart sounds with a digital stethoscope. Innocent murmurs were validated by a pediatric cardiologist and pathologic murmurs were validated by echocardiogram. To augment our collection of normal heart sounds, we utilized a public database of pediatric heart sound recordings (Oliveira, 2022). We propose two novel approaches for this audio classification task. We train a vision transformer on either Markov transition field or Gramian angular field image representations of the frequency spectrum. We benchmark our results against a ResNet-50 CNN trained on spectrogram images. RESULTS: Our final dataset consisted of 366 normal heart sounds, 175 innocent murmurs, and 216 pathologic murmurs. Innocent murmurs collected include Still's murmur, venous hum, and flow murmurs. Pathologic murmurs included ventricular septal defect, tetralogy of Fallot, aortic regurgitation, aortic stenosis, pulmonary stenosis, mitral regurgitation and stenosis, and tricuspid regurgitation. We find that the Vision Transformer consistently outperforms the ResNet-50 on all three image representations, and that the Gramian angular field is the superior image representation for pediatric heart sounds. We calculated a one-vs-rest multi-class ROC curve for each of the three classes. Our best model achieves an area under the curve (AUC) value of 0.92 ± 0.05, 0.83 ± 0.04, and 0.88 ± 0.04 for identifying normal heart sounds, innocent murmurs, and pathologic murmurs, respectively. CONCLUSION: We present two novel methods for pediatric heart sound classification, which outperforms the current standard of using a convolutional neural network trained on spectrogram images. To our knowledge, we are the first to demonstrate multi-class classification of pediatric murmurs. Multiclass output affords a more explainable and interpretable model, which can facilitate further model improvement in the downstream model development cycle and enhance clinician trust and therefore adoption.

A Cohort Study to Evaluate Genetic Predictors for Aromatase Inhibitor Musculoskeletal Symptoms (AIMSS): Results from ECOG-ACRIN E1Z11.

PURPOSE: Aromatase Inhibitor-Associated Musculoskeletal Symptoms (AIMSS) are common and frequently lead to AI discontinuation. Single nucleotide polymorphisms (SNPs) in candidate genes have been associated with AIMSS and AI discontinuation. E1Z11 is a prospective cohort study designed to validate associations between 10 SNPs and AI discontinuation due to AIMSS. PATIENTS AND METHODS: Postmenopausal women with stage I-III hormone receptor-positive breast cancer received anastrozole 1 mg daily and completed patient-reported outcomes (PRO) to assess AIMSS (Stanford Health Assessment Questionnaire; HAQ) at baseline, 3, 6, 9, and 12 months. We estimated that 40% of participants would develop AIMSS, and 25% would discontinue AI treatment within 12 months. Enrollment of 1,000 women with a fixed number per racial strata provided 80% power to detect an effect size of 1.5-4. SNPs were in ESR1 (rs2234693, rs2347868, rs9340835), CYP19A1 (rs1062033, rs4646), TCL1A (rs11849538, rs2369049, rs7158782, rs7159713), and HTR2A (rs2296972). RESULTS: Of 970 evaluable women, 43% developed AIMSS and 12% discontinued AI therapy within 12 months. While more Black and Asian women developed AIMSS compared to White women (49% vs 39%, p=0.017; 50% vs 39%, p=0.004, respectively), AI discontinuation rates were similar across groups. None of the SNPs were significantly associated with AIMSS or AI discontinuation in the overall population, or in distinct cohorts. The odds ratio for rs2296972 (HTR2A) approached significance for developing AIMSS. CONCLUSION: We were unable to prospectively validate candidate SNPs previously associated with AI discontinuation due to AIMSS. Future analyses will explore additional genetic markers, PRO predictors of AIMSS, and differences by race.

Leukemic mutation FLT3-ITD is retained in dendritic cells and disrupts their homeostasis leading to expanded Th17 frequency.

Dendritic cells (DC) are mediators between innate and adaptive immune responses to pathogens and tumors. DC development is determined by signaling through the receptor tyrosine kinase Fms-like tyrosine kinase 3 (FLT3) in bone marrow myeloid progenitors. Recently the naming conventions for DC phenotypes have been updated to distinguish between "Conventional" DCs (cDCs) and plasmacytoid DCs (pDCs). Activating mutations of FLT3, including Internal Tandem Duplication (FLT3-ITD), are associated with poor prognosis for acute myeloid leukemia (AML) patients. Having a shared myeloid lineage it can be difficult to distinguish bone fide DCs from AML tumor cells. To date, there is little information on the effects of FLT3-ITD in DC biology. To further elucidate this relationship we utilized CITE-seq technology in combination with flow cytometry and multiplex immunoassays to measure changes to DCs in human and mouse tissues. We examined the cDC phenotype and frequency in bone marrow aspirates from patients with AML to understand the changes to cDCs associated with FLT3-ITD. When compared to healthy donor (HD) we found that a subset of FLT3-ITD+ AML patient samples have overrepresented populations of cDCs and disrupted phenotypes. Using a mouse model of FLT3-ITD+ AML, we found that cDCs were increased in percentage and number compared to control wild-type (WT) mice. Single cell RNA-seq identified FLT3-ITD+ cDCs as skewed towards a cDC2 T-bet- phenotype, previously shown to promote Th17 T cells. We assessed the phenotypes of CD4+ T cells in the AML mice and found significant enrichment of both Treg and Th17 CD4+ T cells in the bone marrow and spleen compartments. Ex vivo stimulation of CD4+ T cells also showed increased Th17 phenotype in AML mice. Moreover, co-culture of AML mouse-derived DCs and naïve OT-II cells preferentially skewed T cells into a Th17 phenotype. Together, our data suggests that FLT3-ITD+ leukemia-associated cDCs polarize CD4+ T cells into Th17 subsets, a population that has been shown to be negatively associated with survival in solid tumor contexts. This illustrates the complex tumor microenvironment of AML and highlights the need for further investigation into the effects of FLT3-ITD mutations on DC phenotypes and their downstream effects on Th polarization.

Long-Term Follow-Up of the Anthracyclines in Early Breast Cancer Trials (USOR 06-090, NSABP B-46-I/USOR 07132, and NSABP B-49 [NRG Oncology]).

Clinical trials frequently include multiple end points that mature at different times. The initial report, typically based on the primary end point, may be published when key planned co-primary or secondary analyses are not yet available. Clinical Trial Updates provide an opportunity to disseminate additional results from studies, published in JCO or elsewhere, for which the primary end point has already been reported.The primary joint efficacy analysis of the Anthracyclines in Early Breast Cancer (ABC) trials reported in 2017 failed to demonstrate nonanthracycline adjuvant therapy was noninferior to anthracycline-based regimens in high-risk, early breast cancer. Full analyses of the studies had proceeded when the prespecified futility boundary was crossed at a planned futility analysis for the ability to demonstrate noninferiority of a nonanthracycline regimen with continued follow-up. These results were presented with 3.3 years of median follow-up. This manuscript reports results of the final analyses of the study efficacy end points conducted with 6.9 years of median follow-up. Long-term analysis of invasive disease-free survival (IDFS), the primary end point of the ABC trials, remains consistent with the original results, as noninferiority of the nonanthracycline regimens could not be declared on the basis of the original criteria. The secondary end point of recurrence-free interval, which excluded deaths not due to breast cancer as events, favored anthracycline-based regimens, and tests for heterogeneity were significant for hormone receptor status (P = .02) favoring anthracycline regimens for the hormone receptor-negative cohorts. There was no difference in overall survival, and review of the type of IDFS events in the groups suggested reductions in cancer recurrences achieved with anthracycline regimens were offset by late leukemias and deaths unrelated to breast cancer.

Predicting flow cytometry crossmatch results from single-antigen bead testing.

The aim of this study was to devise an algorithm that would predict flow cytometry crossmatch (FCXM) results using single-antigen bead (SAB) mean fluorescent intensity (MFI) levels using samples received through the National External Quality Assurance Scheme (NEQAS) 2B external proficiency testing scheme between 2019 and 2023. A total of 159 serum samples were retrospectively screened using LABScreen Single Antigen Class I and II (SAB), and 40 peripheral blood samples were human leucocyte antigen (HLA) typed with LABType SSO. Donor-specific antibodies were identified for each cell-serum combination tested, and cumulative MFI values were calculated for each test before correlating the screening result with the consensus crossmatch results for this scheme. HLA Class I MFIs were combined to predict the T cell crossmatch. For the B cell crossmatch prediction, two options were considered: (i) HLA Class II MFI values alone and (ii) HLA Class I + Class II MFIs. Receiver operating characteristic analysis was carried out to identify the combined MFI threshold that predicted NEQAS consensus results with the greatest sensitivity and specificity. HLA Class I combined MFI >5000 predicted T cell crossmatch results with 96% sensitivity, 100% specificity, 100% positive predictive value (PPV) and 92% negative predictive value (NPV). For B cell results, HLA Class I + Class II combined MFIs >11,000 gave the best model, showing 97% sensitivity, 82% specificity, 96% PPV and 85% NPV. However, for samples with only HLA Class II sensitization, combined MFIs >13,000 improved the B cell crossmatch predictions: 92% sensitivity, 95% specificity, 96% PPV and 91% NPV. Using this model, combined MFI can be used to predict the immunological risk posed by donor-specific antibodies when it is not possible to carry out an FCXM.

Coronary artery dilation in non-hospitalised children with asymptomatic or mild COVID-19.

INTRODUCTION: Infection with Sars-CoV-2 is known to cause cardiac injury and coronary artery changes in moderate to severe acute COVID-19 and post-acute multisystem inflammatory syndrome in children (MIS-C). However, little is known about the potential for cardiac involvement, in particular coronary artery dilation, in asymptomatic or mild cases of COVID-19. METHODS: A retrospective review of children ≤ 18 years of age with a history of asymptomatic or mild COVID-19 disease who underwent echocardiography after Sars-CoV-2 infection is conducted. Patients were excluded if they had been hospitalised for COVID-19/MIS-C or had a history of cardiac disease that could affect coronary artery dimension. Coronary artery dilation was defined as the Boston Z-score greater than 2.0. RESULTS: One hundred and fifty-seven patients met inclusion criteria with a mean age of 9.4 years (+/- 5.4 years). Eighty-four (54%) patients were identified as having COVID-19 through positive antibody testing. All patients underwent electrocardiogram and echocardiogram as part of their cardiology evaluation. One hundred and thirty-five (86%) patients had a normal evaluation or only a minor variant on electrocardiogram, while 22 patients had abnormalities on echocardiogram, 4 of which demonstrated coronary artery dilation based on the Boston Z-score. CONCLUSIONS: Much of the literature for post-infectious screening and follow-up focuses on patients with a history of moderate to severe COVID-19 disease, emphasising the need for surveillance for the potential development of myocarditis. In this study, 4 out of 157 (2.5%) children with a history of asymptomatic or mild COVID-19 disease without MIS-C were found to have some degree of coronary artery dilation. The significance of this finding currently remains unknown.

Leukemic mutation FLT3-ITD is retained in dendritic cells and disrupts their homeostasis leading to expanded Th17 frequency.

Dendritic cells (DC) are mediators of adaptive immune responses to pathogens and tumors. DC development is determined by signaling through the receptor tyrosine kinase Fms-like tyrosine kinase 3 (FLT3) in bone marrow myeloid progenitors. Recently the naming conventions for DC phenotypes have been updated to distinguish between "Conventional" DCs (cDCs) and plasmacytoid DCs (pDCs). Activating mutations of FLT3, including Internal Tandem Duplication (FLT3-ITD), are associated with poor prognosis for leukemia patients. To date, there is little information on the effects of FLT3-ITD in DC biology. We examined the cDC phenotype and frequency in bone marrow aspirates from patients with acute myeloid leukemia (AML) to understand the changes to cDCs associated with FLT3-ITD. When compared to healthy donor (HD) we found that a subset of FLT3-ITD+ AML patient samples have overrepresented populations of cDCs and disrupted phenotypes. Using a mouse model of FLT3-ITD+ AML, we found that cDCs were increased in percentage and number compared to control wild-type (WT) mice. Single cell RNA-seq identified FLT3-ITD+ cDCs as skewed towards a cDC2 T-bet - phenotype, previously shown to promote Th17 T cells. We assessed the phenotypes of CD4+ T cells in the AML mice and found significant enrichment of both Treg and Th17 CD4+ T cells. Furthermore, co-culture of AML mouse- derived DCs and naïve OT-II cells preferentially skewed T cells into a Th17 phenotype. Together, our data suggests that FLT3-ITD+ leukemia-associated cDCs polarize CD4+ T cells into Th17 subsets, a population that has been shown to be negatively associated with survival in solid tumor contexts. This illustrates the complex tumor microenvironment of AML and highlights the need for further investigation into the effects of FLT3-ITD mutations on DC phenotypes.

Reducing Mode Collapse With Monge-Kantorovich Optimal Transport for Generative Adversarial Networks.

Mode collapse has been a persisting challenge in generative adversarial networks (GANs), and it directly affects the applications of GAN in many domains. Existing works that attempt to solve this problem have some serious limitations: models using optimal transport (OT) strategies (e.g., Wasserstein distance) lead to vanishing or exploding gradients; increasing the number of generators can cause several generators focusing on the same mode; and approaches that modify the loss also do not satisfactorily resolve mode collapse. In this article, we reduce mode collapse by formulating it as a Monge problem of OT map. We show that the Monge problem can be transformed to the distribution transformation problem in GAN, and a rectified affine neural network can be considered as a measurable function. In this way, we propose Monge GAN that uses this measurable function to transform the generated data distribution into the original data distribution. We utilize the Kantorovich formulation to obtain the OT cost, which is regarded as the OT distance between the two distributions. Finally, we conduct extensive experiments on both image and numerical datasets to validate our Monge GAN in reducing model collapse.

Results of a national survey of substance use treatment services for youth under community supervision.

BACKGROUND: Despite the heightened risk for substance use (SU) among youth in the juvenile justice system, many do not receive the treatment that they need. OBJECTIVES: The purpose of this study is to examine the extent to which youth under community supervision by juvenile justice agencies receive community-based SU services and the factors associated with access to such services. METHODS: Data are from a nationally representative sample of Community Supervision (CS) agencies and their primary behavioral health (BH) partners. Surveys were completed by 192 CS and 271 BH agencies. RESULTS: SU services are more often available through BH than CS for all treatment modalities. EBPs are more likely to be used by BH than by CS. Co-location of services occurs most often in communities with fewer treatment options and is associated with higher interagency collaboration. Youth are more likely to receive services in communities with higher EBP use, which mediates the relationship between the availability of SU treatment modalities and the proportion of youth served. CONCLUSION: Findings identify opportunities to strengthen community systems and improve linkage to care.

Quantifying Yeast Microtubules and Spindles Using the Toolkit for Automated Microtubule Tracking (TAMiT).

Fluorescently labeled proteins absorb and emit light, appearing as Gaussian spots in fluorescence imaging. When fluorescent tags are added to cytoskeletal polymers such as microtubules, a line of fluorescence and even non-linear structures results. While much progress has been made in techniques for imaging and microscopy, image analysis is less well-developed. Current analysis of fluorescent microtubules uses either manual tools, such as kymographs, or automated software. As a result, our ability to quantify microtubule dynamics and organization from light microscopy remains limited. Despite the development of automated microtubule analysis tools for in vitro studies, analysis of images from cells often depends heavily on manual analysis. One of the main reasons for this disparity is the low signal-to-noise ratio in cells, where background fluorescence is typically higher than in reconstituted systems. Here, we present the Toolkit for Automated Microtubule Tracking (TAMiT), which automatically detects, optimizes, and tracks fluorescent microtubules in living yeast cells with sub-pixel accuracy. Using basic information about microtubule organization, TAMiT detects linear and curved polymers using a geometrical scanning technique. Images are fit via an optimization problem for the microtubule image parameters that are solved using non-linear least squares in Matlab. We benchmark our software using simulated images and show that it reliably detects microtubules, even at low signal-to-noise ratios. Then, we use TAMiT to measure monopolar spindle microtubule bundle number, length, and lifetime in a large dataset that includes several S. pombe mutants that affect microtubule dynamics and bundling. The results from the automated analysis are consistent with previous work and suggest a direct role for CLASP/Cls1 in bundling spindle microtubules. We also illustrate automated tracking of single curved astral microtubules in S. cerevisiae, with measurement of dynamic instability parameters. The results obtained with our fully-automated software are similar to results using hand-tracked measurements. Therefore, TAMiT can facilitate automated analysis of spindle and microtubule dynamics in yeast cells.

Quantifying yeast microtubules and spindles using the Toolkit for Automated Microtubule Tracking (TAMiT).

Fluorescently labeled proteins absorb and emit light, appearing as Gaussian spots in fluorescence imaging. When fluorescent tags are added to cytoskeletal polymers such as microtubules, a line of fluorescence and even non-linear structures results. While much progress has been made in techniques for imaging and microscopy, image analysis is less well developed. Current analysis of fluorescent microtubules uses either manual tools, such as kymographs, or automated software. As a result, our ability to quantify microtubule dynamics and organization from light microscopy remains limited. Despite development of automated microtubule analysis tools for in vitro studies, analysis of images from cells often depends heavily on manual analysis. One of the main reasons for this disparity is the low signal-to-noise ratio in cells, where background fluorescence is typically higher than in reconstituted systems. Here, we present the Toolkit for Automated Microtubule Tracking (TAMiT), which automatically detects, optimizes and tracks fluorescent microtubules in living yeast cells with sub-pixel accuracy. Using basic information about microtubule organization, TAMiT detects linear and curved polymers using a geometrical scanning technique. Images are fit via an optimization problem for the microtubule image parameters that is solved using non-linear least squares in Matlab. We benchmark our software using simulated images and show that it reliably detects microtubules, even at low signal-to-noise ratios. Then, we use TAMiT to measure monopolar spindle microtubule bundle number, length, and lifetime in a large dataset that includes several S. pombe mutants that affect microtubule dynamics and bundling. The results from the automated analysis are consistent with previous work, and suggest a direct role for CLASP/Cls1 in bundling spindle microtubules. We also illustrate automated tracking of single curved astral microtubules in S. cerevisiae , with measurement of dynamic instability parameters. The results obtained with our fully-automated software are similar to results using hand-tracked measurements. Therefore, TAMiT can facilitate automated analysis of spindle and microtubule dynamics in yeast cells.

Annexin A2 Loss After Cardiopulmonary Bypass and Development of Acute Postoperative Respiratory Dysfunction in Children.

The primary objective of this study was to determine whether expression of the multifunctional and adherens junction-regulating protein, annexin A2 (A2), is altered following cardiopulmonary bypass (CPB). A secondary objective was to determine whether depletion of A2 is associated with post-CPB organ dysfunction in children. DESIGN: In a prospective, observational study conducted over a 1-year period in children undergoing cardiac surgery requiring CPB, we analyzed A2 expression in peripheral blood mononuclear cells at different time points. We then assessed the relationship of A2 expression with organ function at each time point in the early postoperative period. SETTING: Twenty-three-bed mixed PICU in a tertiary academic center. PARTICIPANTS: Patients 1 month to 18 years old undergoing cardiac surgery requiring CPB. MEAN OUTCOME MEASUREMENTS AND RESULTS: We analyzed A2 expression in 22 enrolled subjects (n = 9, 1-23 mo old; n = 13, 2-18 yr old) and found a proteolysis-mediated decline in intact A2 immediately after bypass (p = 0.0009), reaching a median of 4% of baseline at 6 hours after bypass (p < 0.0001), and recovery by postoperative day 1. The degree of A2 depletion immediately after bypass in 1-23-month-olds correlated strongly with the extent of organ dysfunction, as measured by PICU admission Vasoactive-Ventilation-Renal (p = 0.004) and PEdiatric Logistic Organ Dysfunction-2 (p = 0.039) scores on postoperative day 1. A2 depletion immediately after bypass also correlated with more protracted requirement for both respiratory support (p = 0.007) and invasive ventilation (p = 0.013) in the 1-23-month-olds. CONCLUSIONS AND RELEVANCE: The degree of depletion of A2 following CPB correlates with more severe organ dysfunction, especially acute respiratory compromise in children under 2 years. These findings suggest that loss of A2 may contribute to pulmonary microvascular leak in young children following CPB.

BET inhibitors rescue anti-PD1 resistance by enhancing TCF7 accessibility in leukemia-derived terminally exhausted CD8+ T cells.

Many acute myeloid leukemia (AML) patients exhibit hallmarks of immune exhaustion, such as increased myeloid-derived suppressor cells, suppressive regulatory T cells and dysfunctional T cells. Similarly, we have identified the same immune-related features, including exhausted CD8+ T cells (TEx) in a mouse model of AML. Here we show that inhibitors that target bromodomain and extra-terminal domain (BET) proteins affect tumor-intrinsic factors but also rescue T cell exhaustion and ICB resistance. Ex vivo treatment of cells from AML mice and AML patients with BET inhibitors (BETi) reversed CD8+ T cell exhaustion by restoring proliferative capacity and expansion of the more functional precursor-exhausted T cells. This reversal was enhanced by combined BETi and anti-PD1 treatment. BETi synergized with anti-PD1 in vivo, resulting in the reduction of circulating leukemia cells, enrichment of CD8+ T cells in the bone marrow, and increase in expression of Tcf7, Slamf6, and Cxcr5 in CD8+ T cells. Finally, we profiled the epigenomes of in vivo JQ1-treated AML-derived CD8+ T cells by single-cell ATAC-seq and found that JQ1 increases Tcf7 accessibility specifically in Tex cells, suggesting that BETi likely acts mechanistically by relieving repression of progenitor programs in Tex CD8+ T cells and maintaining a pool of anti-PD1 responsive CD8+ T cells.

An early evaluation of the HISTO SPOT® AB ID Class I & II test in cardiothoracic transplant patients.

The HISTO SPOT® AB ID assay (BAG Diagnostics GmbH) is a novel single antigen HLA Class I & II antibody definition test used with the MR.SPOT® processor. We compared this assay with Luminex® -based assays to assess its potential application in defining unacceptable antigens for transplantation in patients awaiting transplants with cardiothoracic organs. A cohort of 40 sensitized cardiothoracic patients were identified, and one sample was selected from each patient. The required screening was based on the patients' antibody profiles (Class I, n = 17, Class II, n = 11, Class I & II, n = 12). Samples were screened with LABScreen™ Single Antigen (SAg), LIFECODES® LSA™, HISTO SPOT® AB ID, and an acid modified LABScreen™ SAg test for detecting antibodies against denatured HLA. Results indicated that HISTO SPOT® AB ID had reduced sensitivity (68% for Class I; 69% for Class II). When compared to LABScreen™ and LIFECODES® , HISTO SPOT® AB ID failed to detect Luminex® -defined antibodies with median fluorescence intensity (MFI) ranging from 1114 to 24,489. The HISTO SPOT® AB ID panel used in the study had reduced antigen representation compared with Luminex® -based assays which further compromised its capacity for antibody detection and definition. Further work is needed to evaluate the clinical relevance of these differences between the performance of HISTO SPOT® and Luminex® -based methods.

The Application of Differing Machine Learning Algorithms and Their Related Performance in Detecting Skin Cancers and Melanomas.

Skin cancer, and its less common form melanoma, is a disease affecting a wide variety of people. Since it is usually detected initially by visual inspection, it makes for a good candidate for the application of machine learning. With early detection being key to good outcomes, any method that can enhance the diagnostic accuracy of dermatologists and oncologists is of significant interest. When comparing different existing implementations of machine learning against public datasets and several we seek to create, we attempted to create a more accurate model that can be readily adapted to use in clinical settings. We tested combinations of models, including convolutional neural networks (CNNs), and various layers of data manipulation, such as the application of Gaussian functions and trimming of images to improve accuracy. We also created more traditional data models, including support vector classification, K-nearest neighbor, Naïve Bayes, random forest, and gradient boosting algorithms, and compared them to the CNN-based models we had created. Results had indicated that CNN-based algorithms significantly outperformed other data models we had created. Partial results of this work were presented at the CSET Presentations for Research Month at the Minnesota State University, Mankato.

Unit cohesion during deployment and post-deployment mental health: is cohesion an individual- or unit-level buffer for combat-exposed soldiers?

BACKGROUND: Unit cohesion may protect service member mental health by mitigating effects of combat exposure; however, questions remain about the origins of potential stress-buffering effects. We examined buffering effects associated with two forms of unit cohesion (peer-oriented horizontal cohesion and subordinate-leader vertical cohesion) defined as either individual-level or aggregated unit-level variables. METHODS: Longitudinal survey data from US Army soldiers who deployed to Afghanistan in 2012 were analyzed using mixed-effects regression. Models evaluated individual- and unit-level interaction effects of combat exposure and cohesion during deployment on symptoms of post-traumatic stress disorder (PTSD), depression, and suicidal ideation reported at 3 months post-deployment (model n's = 6684 to 6826). Given the small effective sample size (k = 89), the significance of unit-level interactions was evaluated at a 90% confidence level. RESULTS: At the individual-level, buffering effects of horizontal cohesion were found for PTSD symptoms [B = -0.11, 95% CI (-0.18 to -0.04), p < 0.01] and depressive symptoms [B = -0.06, 95% CI (-0.10 to -0.01), p < 0.05]; while a buffering effect of vertical cohesion was observed for PTSD symptoms only [B = -0.03, 95% CI (-0.06 to -0.0001), p < 0.05]. At the unit-level, buffering effects of horizontal (but not vertical) cohesion were observed for PTSD symptoms [B = -0.91, 90% CI (-1.70 to -0.11), p = 0.06], depressive symptoms [B = -0.83, 90% CI (-1.24 to -0.41), p < 0.01], and suicidal ideation [B = -0.32, 90% CI (-0.62 to -0.01), p = 0.08]. CONCLUSIONS: Policies and interventions that enhance horizontal cohesion may protect combat-exposed units against post-deployment mental health problems. Efforts to support individual soldiers who report low levels of horizontal or vertical cohesion may also yield mental health benefits.

Chromatin bridges, not micronuclei, activate cGAS after drug-induced mitotic errors in human cells.

Mitotic errors can activate cyclic GMP-AMP synthase (cGAS) and induce type I interferon (IFN) signaling. Current models propose that chromosome segregation errors generate micronuclei whose rupture activates cGAS. We used a panel of antimitotic drugs to perturb mitosis in human fibroblasts and measured abnormal nuclear morphologies, cGAS localization, and IFN signaling in the subsequent interphase. Micronuclei consistently recruited cGAS without activating it. Instead, IFN signaling correlated with formation of cGAS-coated chromatin bridges that were selectively generated by microtubule stabilizers and MPS1 inhibitors. cGAS activation by chromatin bridges was suppressed by drugs that prevented cytokinesis. We confirmed cGAS activation by chromatin bridges in cancer lines that are unable to secrete IFN by measuring paracrine transfer of 2'3'-cGAMP to fibroblasts, and in mouse cells. We propose that cGAS is selectively activated by self-chromatin when it is stretched in chromatin bridges. Immunosurveillance of cells that fail mitosis, and antitumor actions of taxanes and MPS1 inhibitors, may depend on this effect.

Fast Local Spatial Verification for Feature-Agnostic Large-Scale Image Retrieval.

Images from social media can reflect diverse viewpoints, heated arguments, and expressions of creativity, adding new complexity to retrieval tasks. Researchers working on Content-Based Image Retrieval (CBIR) have traditionally tuned their algorithms to match filtered results with user search intent. However, we are now bombarded with composite images of unknown origin, authenticity, and even meaning. With such uncertainty, users may not have an initial idea of what the search query results should look like. For instance, hidden people, spliced objects, and subtly altered scenes can be difficult for a user to detect initially in a meme image, but may contribute significantly to its composition. It is pertinent to design systems that retrieve images with these nuanced relationships in addition to providing more traditional results, such as duplicates and near-duplicates - and to do so with enough efficiency at large scale. We propose a new approach for spatial verification that aims at modeling object-level regions using image keypoints retrieved from an image index, which is then used to accurately weight small contributing objects within the results, without the need for costly object detection steps. We call this method the Objects in Scene to Objects in Scene (OS2OS) score, and it is optimized for fast matrix operations, which can run quickly on either CPUs or GPUs. It performs comparably to state-of-the-art methods on classic CBIR problems (Oxford 5K, Paris 6K, and Google-Landmarks), and outperforms them in emerging retrieval tasks such as image composite matching in the NIST MFC2018 dataset and meme-style imagery from Reddit.