Climate and Gender: Association Between Droughts and Intimate Partner Violence in India.

Extreme climate events are related to women's exposure to different forms of violence. We examined the relationship between droughts and physical, sexual, and emotional intimate partner violence (IPV) in India by using 2 different definitions of drought: precipitation-based drought and socioeconomic drought. We analyzed data from 2 rounds of a nationally representative survey, the National Family Health Survey, where married women were asked about their experiences of IPV in the previous year (2015-2016 and 2019-2021; n = 122,696). Precipitation-based drought was estimated using remote sensing data and geographic information system (GIS) mapping, while socioeconomic drought status was collected from government records. Logistic regression models showed precipitation-based drought to increase the risk of experiencing physical IPV and emotional IPV. Similar findings were observed for socioeconomic drought; women residing in areas classified as drought-impacted by the government were more likely to report physical IPV, sexual IPV, and emotional IPV. These findings support the growing body of evidence regarding the relationship between climate change and women's vulnerability, and highlight the need for gender responsive strategies for disaster management and preparedness.

Plug-and-Play Model-Agnostic Counterfactual Policy Synthesis for Deep Reinforcement Learning-Based Recommendation.

Recent advances in recommender systems have proved the potential of reinforcement learning (RL) to handle the dynamic evolution processes between users and recommender systems. However, learning to train an optimal RL agent is generally impractical with commonly sparse user feedback data in the context of recommender systems. To circumvent the lack of interaction of current RL-based recommender systems, we propose to learn a general model-agnostic counterfactual synthesis (MACS) policy for counterfactual user interaction data augmentation. The counterfactual synthesis policy aims to synthesize counterfactual states while preserving significant information in the original state relevant to the user's interests, building upon two different training approaches we designed: learning with expert demonstrations and joint training. As a result, the synthesis of each counterfactual data is based on the current recommendation agent's interaction with the environment to adapt to users' dynamic interests. We integrate the proposed policy deep deterministic policy gradient (DDPG), soft actor critic (SAC), and twin delayed DDPG (TD3) in an adaptive pipeline with a recommendation agent that can generate counterfactual data to improve the performance of recommendation. The empirical results on both online simulation and offline datasets demonstrate the effectiveness and generalization of our counterfactual synthesis policy and verify that it improves the performance of RL recommendation agents.

Robust Multi-View Fracture Detection in the Presence of Other Abnormalities Using HAMIL-Net.

INTRODUCTION: Foot and ankle fractures are the most common military health problem. Automated diagnosis can save time and personnel. It is crucial to distinguish fractures not only from normal healthy cases, but also robust against the presence of other orthopedic pathologies. Artificial intelligence (AI) deep learning has been shown to be promising. Previously, we have developed HAMIL-Net to automatically detect orthopedic injuries for upper extremity injuries. In this research, we investigated the performance of HAMIL-Net for detecting foot and ankle fractures in the presence of other abnormalities. MATERIALS AND METHODS: HAMIL-Net is a novel deep neural network consisting of a hierarchical attention layer followed by a multiple-instance learning layer. The design allowed it to deal with imaging studies with multiple views. We used 148K musculoskeletal imaging studies for 51K Veterans at VA San Diego in the past 20 years to create datasets for this research. We annotated each study by a semi-automated pipeline leveraging radiology reports written by board-certified radiologists and extracting findings with a natural language processing tool and manually validated the annotations. RESULTS: HAMIL-Net can be trained with study-level, multiple-view examples, and detect foot and ankle fractures with a 0.87 area under the receiver operational curve, but the performance dropped when tested by cases including other abnormalities. By integrating a fracture specialized model with one that detecting a broad range of abnormalities, HAMIL-Net's accuracy of detecting any abnormality improved from 0.53 to 0.77 and F-score from 0.46 to 0.86. We also reported HAMIL-Net's performance under different study types including for young (age 18-35) patients. CONCLUSIONS: Automated fracture detection is promising but to be deployed in clinical use, presence of other abnormalities must be considered to deliver its full benefit. Our results with HAMIL-Net showed that considering other abnormalities improved fracture detection and allowed for incidental findings of other musculoskeletal abnormalities pertinent or superimposed on fractures.

Violence against women on Twitter in India: Testing a taxonomy for online misogyny and measuring its prevalence during COVID-19.

BACKGROUND: Online misogyny is a violation of women's digital rights. Empirical studies on this topic are however lacking, particularly in low- and middle- income countries. The current study aimed to estimate whether prevalence of online misogyny on Twitter in India changed since the pandemic. METHODS: Based on prior theoretical work, we defined online misogyny as consisting of six overlapping forms: sexist abuses, sexual objectification, threatening to physically or sexually harm women, asserting women's inferiority, justifying violence against women, and dismissing feminist efforts. Qualitative analysis of a small subset of tweets posted from India (40,672 tweets) substantiated this definition and taxonomy for online misogyny. Supervised machine learning models were used to predict the status of misogyny across a corpus of 30 million tweets posted from India between 2018 and 2021. Next, interrupted time series analysis examined changes in online misogyny prevalence, before and during COVID-19. RESULTS: Qualitative assessment showed that online misogyny in India existed most in the form of sexual objectification and sexist abusive content, which demeans women and shames them for their presumed sexual activity. Around 2% of overall tweets posted from India between 2018 and 2021 included some form of misogynistic content. The absolute volume as well as proportion of misogynistic tweets showed significant increasing trends after the onset of COVID-19, relative to trends prior to the pandemic. CONCLUSION: Findings highlight increasing gender inequalities on Twitter since the pandemic. Aggressive and hateful tweets that target women attempt to reinforce traditional gender norms, especially those relating to idealized sexual behavior and framing of women as sexual beings. There is an urgent need for future research and development of interventions to make digital spaces gender equitable and welcoming to women.

How large language models can augment perioperative medicine: a daring discourse.

Interest in natural language processing, specifically large language models, for clinical applications has exploded in a matter of several months since the introduction of ChatGPT. Large language models are powerful and impressive. It is important that we understand the strengths and limitations of this rapidly evolving technology so that we can brainstorm its future potential in perioperative medicine. In this daring discourse, we discuss the issues with these large language models and how we should proactively think about how to leverage these models into practice to improve patient care, rather than worry that it may take over clinical decision-making. We review three potential major areas in which it may be used to benefit perioperative medicine: (1) clinical decision support and surveillance tools, (2) improved aggregation and analysis of research data related to large retrospective studies and application in predictive modeling, and (3) optimized documentation for quality measurement, monitoring and billing compliance. These large language models are here to stay and, as perioperative providers, we can either adapt to this technology or be curtailed by those who learn to use it well.

Using machine learning to understand determinants of IUD use in India: Analyses of the National Family Health Surveys (NFHS-4).

Intra-uterine devices (IUDs) are a safe and effective method to delay or space pregnancies and are available for free or at low cost in the Indian public health system; yet, IUD uptake in India remains low. Limited quantitative research using national data has explored factors that may affect IUD use. Machine Learning (ML) techniques allow us to explore determinants of low prevalence behaviors in survey research, such as IUD use. We applied ML to explore the determinants of IUD use in India among married women in the 4th National Family Health Survey (NFHS-4; N = 499,627), which collects data on demographic and health indicators among women of childbearing age. We conducted ML logistic regression (lasso and ridge) and neural network approaches to assess significant determinants and used iterative thematic analysis (ITA) to offer insight into related variable constructs generated from a series of regularized models. We found that couples' shared family planning (FP) goals were the strongest determinants of IUD use, followed by receipt of FP services and desire for no more children, higher wealth and education, and receipt of maternal and child health services. Findings highlight the importance of male engagement and family planning services for IUD uptake and the need for more targeted efforts to support awareness of IUD as an option for spacing, especially for those of lower SES and with lower access to care.

RadBERT: Adapting Transformer-based Language Models to Radiology.

Purpose: To investigate if tailoring a transformer-based language model to radiology is beneficial for radiology natural language processing (NLP) applications. Materials and Methods: This retrospective study presents a family of bidirectional encoder representations from transformers (BERT)-based language models adapted for radiology, named RadBERT. Transformers were pretrained with either 2.16 or 4.42 million radiology reports from U.S. Department of Veterans Affairs health care systems nationwide on top of four different initializations (BERT-base, Clinical-BERT, robustly optimized BERT pretraining approach [RoBERTa], and BioMed-RoBERTa) to create six variants of RadBERT. Each variant was fine-tuned for three representative NLP tasks in radiology: (a) abnormal sentence classification: models classified sentences in radiology reports as reporting abnormal or normal findings; (b) report coding: models assigned a diagnostic code to a given radiology report for five coding systems; and (c) report summarization: given the findings section of a radiology report, models selected key sentences that summarized the findings. Model performance was compared by bootstrap resampling with five intensively studied transformer language models as baselines: BERT-base, BioBERT, Clinical-BERT, BlueBERT, and BioMed-RoBERTa. Results: For abnormal sentence classification, all models performed well (accuracies above 97.5 and F1 scores above 95.0). RadBERT variants achieved significantly higher scores than corresponding baselines when given only 10% or less of 12 458 annotated training sentences. For report coding, all variants outperformed baselines significantly for all five coding systems. The variant RadBERT-BioMed-RoBERTa performed the best among all models for report summarization, achieving a Recall-Oriented Understudy for Gisting Evaluation-1 score of 16.18 compared with 15.27 by the corresponding baseline (BioMed-RoBERTa, P < .004). Conclusion: Transformer-based language models tailored to radiology had improved performance of radiology NLP tasks compared with baseline transformer language models.Keywords: Translation, Unsupervised Learning, Transfer Learning, Neural Networks, Informatics Supplemental material is available for this article. © RSNA, 2022See also commentary by Wiggins and Tejani in this issue.

Rethink, Revisit, Revise: A Spiral Reinforced Self-Revised Network for Zero-Shot Learning.

Current approaches to zero-shot learning (ZSL) struggle to learn generalizable semantic knowledge capable of capturing complex correlations. Inspired by Spiral Curriculum, which enhances learning processes by revisiting knowledge, we propose a form of spiral learning that revisits visual representations based on a sequence of attribute groups (e.g., a combined group of color and shape). Spiral learning aims to learn generalized local correlations, enabling models to gradually enhance global learning and, thus, understand complex correlations. Our implementation is based on a two-stage reinforced self-revised (RSR) framework: preview and review. RSR first previews visual information to construct diverse attribute groups in a weakly supervised manner. Then, it spirally learns refined localities based on attribute groups and uses localities to revise global semantic correlations. Our framework outperforms state-of-the-art algorithms on four benchmark datasets in both zero-shot and generalized zero-shot settings, which demonstrates the effectiveness of spiral learning in learning generalizable and complex correlations. We also conduct extensive analysis to show that attribute groups and reinforced decision processes can capture complementary semantic information to improve predictions and aid explainability.

Help seeking behavior by women experiencing intimate partner violence in india: A machine learning approach to identifying risk factors.

BACKGROUND: Despite the low prevalence of help-seeking behavior among victims of intimate partner violence (IPV) in India, quantitative evidence on risk factors, is limited. We use a previously validated exploratory approach, to examine correlates of help-seeking from anyone (e.g. family, friends, police, doctor etc.), as well as help-seeking from any formal sources. METHODS: We used data from a nationally-representative health survey conducted in 2015-16 in India, and included all variables in the dataset (~6000 variables) as independent variables. Two machine learning (ML) models were used- L-1, and L-2 regularized logistic regression models. The results from these models were qualitatively coded by researchers to identify broad themes associated with help-seeking behavior. This process of implementing ML models followed by qualitative coding was repeated until pre-specified criteria were met. RESULTS: Identified themes associated with help-seeking behavior included experience of injury from violence, husband's controlling behavior, husband's consumption of alcohol, and being currently separated from husband. Themes related to women's access to social and economic resources, such as women's employment, and receipt of maternal and reproductive health services were also noted to be related factors. We observed similarity in correlates for seeking help from anyone, vs from formal sources, with a greater focus on women being separated for help-seeking from formal sources. CONCLUSION: Findings highlight the need for community programs to reach out to women trapped in abusive relationships, as well as the importance of women's social and economic connectedness; future work should consider holistic interventions that integrate IPV screening and support services with women's health related services.

Machine learning analysis of non-marital sexual violence in India.

BACKGROUND: Machine learning techniques can explore low prevalence data to offer insight into identification of factors associated with non-marital sexual violence (NMSV). NMSV in India is a health and human rights concern that disproportionately affects adolescents, is under-reported, and not well understood or addressed in the country. METHODS: We applied machine learning methods to retrospective cross-sectional data from India's nationally-representative National Family Health Survey 4, a demographic and health study conducted in 2015-16, which offers 4000+ variables as potential independent variables. We used Least Absolute Shrinkage and Selection Operator (lasso) or L-1 regularized logistic regression models as well as L-2 regularized logistic regression or ridge models; we conducted an iterative thematic analysis (ITA) of variables generated from a series of regularized models. FINDINGS: Thematic analysis of regularized models highlight that past exposure to violence was most predictive of NMSV, followed by geography, sexual behavior, and poor sexual and reproductive health knowledge. After these, indicators largely related to resources and autonomy (e.g., access to health services, and income generating) were associated with NMSV. Exploratory analysis with the subsample of never married adolescents 15-19 years old, a population with higher representation of recent NMSV, further emphasized the role of wealth and mobility as key correlates of NMSV, along with poor HIV knowledge, tobacco use, higher fertility preferences, and attitudes accepting of marital violence. INTERPRETATION: Findings indicate the validity of machine learning with iterative theme analysis (ITA) to identify factors associated with violence. Findings were consistent with prior work demonstrating associations between NMSV and other violence experiences, but also showed novel correlates such as lower SRH knowledge and service utilization and, for girls, norms and preferences suggesting more restrictive gender norms. Sexual and reproductive health, gender equity and safety focused interventions are important for addressing NMSV in India, particularly for adolescents.

Application of machine learning to understand child marriage in India.

BACKGROUND: Prior research documents that India has the greatest number of girls married as minors of any nation in the world, increasing social and health risks for both these young wives and their children. While the prevalence of child marriage has declined in the nation, more work is needed to accelerate this decline and the negative consequences of the practice. Expanded targets for intervention require greater identification of these targets. Machine learning can offer insight into identification of novel factors associated with child marriage that can serve as targets for intervention. METHODS: We applied machine learning methods to retrospective cross-sectional survey data from India on demographics and health, the nationally-representative National Family Health Survey, conducted in 2015-16. We analyzed data using a traditional regression model, with child marriage as the dependent variable, and 4000+ variables from the survey as the independent variables. We also used three commonly used machine learning algorithms- Least Absolute Shrinkage and Selection Operator (lasso) or L-1 regularized logistic regression models; L2 regularized logistic regression or ridge models; and neural network models. Finally, we developed and applied a novel and rigorous approach involving expert qualitative review and coding of variables generated from an iterative series of regularized models to assess thematically key variable groupings associated with child marriage. FINDINGS: Analyses revealed that regularized logistic and neural network applications demonstrated better accuracy and lower error rates than traditional logistic regression, with a greater number of features and variables generated. Regularized models highlight higher fertility and contraception, longer duration of marriage, geographic, and socioeconomic vulnerabilities as key correlates; findings shown in prior research. However, our novel method involving expert qualitative coding of variables generated from iterative regularized models and resultant thematic generation offered clarity on variables not focused upon in prior research, specifically non-utilization of health system benefits related to nutrition for mothers and infants. INTERPRETATION: Machine learning appears to be a valid means of identifying key correlates of child marriage in India and, via our innovative iterative thematic approach, can be useful to identify novel variables associated with this outcome. Findings related to low nutritional service uptake also demonstrate the need for more focus on public health outreach for nutritional programs tailored to this population.

The Performance of an Artificial Neural Network Model in Predicting the Early Distribution Kinetics of Propofol in Morbidly Obese and Lean Subjects.

BACKGROUND: Induction of anesthesia is a phase characterized by rapid changes in both drug concentration and drug effect. Conventional mammillary compartmental models are limited in their ability to accurately describe the early drug distribution kinetics. Recirculatory models have been used to account for intravascular mixing after drug administration. However, these models themselves may be prone to misspecification. Artificial neural networks offer an advantage in that they are flexible and not limited to a specific structure and, therefore, may be superior in modeling complex nonlinear systems. They have been used successfully in the past to model steady-state or near steady-state kinetics, but never have they been used to model induction-phase kinetics using a high-resolution pharmacokinetic dataset. This study is the first to use an artificial neural network to model early- and late-phase kinetics of a drug. METHODS: Twenty morbidly obese and 10 lean subjects were each administered propofol for induction of anesthesia at a rate of 100 mg/kg/h based on lean body weight and total body weight for obese and lean subjects, respectively. High-resolution plasma samples were collected during the induction phase of anesthesia, with the last sample taken at 16 hours after propofol administration for a total of 47 samples per subject. Traditional mammillary compartment models, recirculatory models, and a gated recurrent unit neural network were constructed to model the propofol pharmacokinetics. Model performance was compared. RESULTS: A 4-compartment model, a recirculatory model, and a gated recurrent unit neural network were assessed. The final recirculatory model (mean prediction error: 0.348; mean square error: 23.92) and gated recurrent unit neural network that incorporated ensemble learning (mean prediction error: 0.161; mean square error: 20.83) had similar performance. Each of these models overpredicted propofol concentrations during the induction and elimination phases. Both models had superior performance compared to the 4-compartment model (mean prediction error: 0.108; mean square error: 31.61), which suffered from overprediction bias during the first 5 minutes followed by under-prediction bias after 5 minutes. CONCLUSIONS: A recirculatory model and gated recurrent unit artificial neural network that incorporated ensemble learning both had similar performance and were both superior to a compartmental model in describing our high-resolution pharmacokinetic data of propofol. The potential of neural networks in pharmacokinetic modeling is encouraging but may be limited by the amount of training data available for these models.

Identifying and characterizing highly similar notes in big clinical note datasets.

BACKGROUND: Big clinical note datasets found in electronic health records (EHR) present substantial opportunities to train accurate statistical models that identify patterns in patient diagnosis and outcomes. However, near-to-exact duplication in note texts is a common issue in many clinical note datasets. We aimed to use a scalable algorithm to de-duplicate notes and further characterize the sources of duplication. METHODS: We use an approximation algorithm to minimize pairwise comparisons consisting of three phases: (1) Minhashing with Locality Sensitive Hashing; (2) a clustering method using tree-structured disjoint sets; and (3) classification of near-duplicates (exact copies, common machine output notes, or similar notes) via pairwise comparison of notes in each cluster. We use the Jaccard Similarity (JS) to measure similarity between two documents. We analyzed two big clinical note datasets: our institutional dataset and MIMIC-III. RESULTS: There were 1,528,940 notes analyzed from our institution. The de-duplication algorithm completed in 36.3 h. When the JS threshold was set at 0.7, the total number of clusters was 82,371 (total notes = 304,418). Among all JS thresholds, no clusters contained pairs of notes that were incorrectly clustered. When the JS threshold was set at 0.9 or 1.0, the de-duplication algorithm captured 100% of all random pairs with their JS at least as high as the set thresholds from the validation set. Similar performance was noted when analyzing the MIMIC-III dataset. CONCLUSIONS: We showed that among the EHR from our institution and from the publicly-available MIMIC-III dataset, there were a significant number of near-to-exact duplicated notes.

Fast Inference with Min-Sum Matrix Product.

The MAP inference problem in many graphical models can be solved efficiently using a fast algorithm for computing min-sum products of n × n matrices. The class of models in question includes cyclic and skip-chain models that arise in many applications. Although the worst-case complexity of the min-sum product operation is not known to be much better than O(n(3)), an O(n(2.5)) expected time algorithm was recently given, subject to some constraints on the input matrices. In this paper, we give an algorithm that runs in O(n(2) log n) expected time, assuming that the entries in the input matrices are independent samples from a uniform distribution. We also show that two variants of our algorithm are quite fast for inputs that arise in several applications. This leads to significant performance gains over previous methods in applications within computer vision and natural language processing.

Faster graphical models for point-pattern matching.

It has been shown that isometric matching problems can be solved exactly in polynomial time, by means of a Junction Tree with small maximal clique size. Recently, an iterative algorithm was presented which converges to the same solution an order of magnitude faster. Here, we build on both of these ideas to produce an algorithm with the same asymptotic running time as the iterative solution, but which requires only a single iteration of belief propagation. Thus our algorithm is much faster in practice, while maintaining similar error rates.

Learning graph matching.

As a fundamental problem in pattern recognition, graph matching has applications in a variety of fields, from computer vision to computational biology. In graph matching, patterns are modeled as graphs and pattern recognition amounts to finding a correspondence between the nodes of different graphs. Many formulations of this problem can be cast in general as a quadratic assignment problem, where a linear term in the objective function encodes node compatibility and a quadratic term encodes edge compatibility. The main research focus in this theme is about designing efficient algorithms for approximately solving the quadratic assignment problem, since it is NP-hard. In this paper we turn our attention to a different question: how to estimate compatibility functions such that the solution of the resulting graph matching problem best matches the expected solution that a human would manually provide. We present a method for learning graph matching: the training examples are pairs of graphs and the 'labels' are matches between them. Our experimental results reveal that learning can substantially improve the performance of standard graph matching algorithms. In particular, we find that simple linear assignment with such a learning scheme outperforms Graduated Assignment with bistochastic normalisation, a state-of-the-art quadratic assignment relaxation algorithm.

Graph rigidity, cyclic belief propagation, and point pattern matching.

A recent paper [1] proposed a provably optimal polynomial time method for performing near-isometric point pattern matching by means of exact probabilistic inference in a chordal graphical model. Its fundamental result is that the chordal graph in question is shown to be globally rigid, implying that exact inference provides the same matching solution as exact inference in a complete graphical model. This implies that the algorithm is optimal when there is no noise in the point patterns. In this paper, we present a new graph that is also globally rigid but has an advantage over the graph proposed in [1]: Its maximal clique size is smaller, rendering inference significantly more efficient. However, this graph is not chordal, and thus, standard Junction Tree algorithms cannot be directly applied. Nevertheless, we show that loopy belief propagation in such a graph converges to the optimal solution. This allows us to retain the optimality guarantee in the noiseless case, while substantially reducing both memory requirements and processing time. Our experimental results show that the accuracy of the proposed solution is indistinguishable from that in [1] when there is noise in the point patterns.