RESUMO
BACKGROUND: In recent years, numerous methods have been introduced to predict glucose levels using machine-learning techniques on patients' daily behavioral and continuous glucose data. Nevertheless, a definitive consensus remains elusive regarding modeling the combined effects of diet and exercise for optimal glucose prediction. A notable challenge is the propensity for observational patient datasets from uncontrolled environments to overfit due to skewed feature distributions of target behaviors; for instance, diabetic patients seldom engage in high-intensity exercise post-meal. METHODS: In this study, we introduce a unique application of Bayesian transfer learning for postprandial glucose prediction using randomized controlled trial (RCT) data. The data comprises a time series of three key variables: continuous glucose levels, exercise expenditure, and carbohydrate intake. For building the optimal model to predict postprandial glucose levels we initially gathered balanced training data from RCTs on healthy participants by randomizing behavioral conditions. Subsequently, we pretrained the model's parameter distribution using RCT data from the healthy cohort. This pretrained distribution was then adjusted, transferred, and utilized to determine the model parameters for each patient. RESULTS: The efficacy of the proposed method was appraised using data from 68 gestational diabetes mellitus (GDM) patients in uncontrolled settings. The evaluation underscored the enhanced performance attained through our method. Furthermore, when modeling the joint impact of diet and exercise, the synergetic model proved more precise than its additive counterpart. CONCLUSION: An innovative application of the transfer-learning utilizing randomized controlled trial data can improve the challenging modeling task of postprandial glucose prediction for GDM patients, integrating both dietary and exercise behaviors. For more accurate prediction, future research should focus on incorporating the long-term effects of exercise and other glycemic-related factors such as stress, sleep.
Assuntos
Teorema de Bayes , Glicemia , Exercício Físico , Aprendizado de Máquina , Período Pós-Prandial , Humanos , Glicemia/metabolismo , Glicemia/análise , Feminino , Período Pós-Prandial/fisiologia , Exercício Físico/fisiologia , Dieta , Gravidez , Diabetes Gestacional/sangue , Diabetes Gestacional/metabolismo , AdultoRESUMO
Methods of estimating polygenic scores (PGSs) from genome-wide association studies are increasingly utilized. However, independent method evaluation is lacking, and method comparisons are often limited. Here, we evaluate polygenic scores derived via seven methods in five biobank studies (totaling about 1.2 million participants) across 16 diseases and quantitative traits, building on a reference-standardized framework. We conducted meta-analyses to quantify the effects of method choice, hyperparameter tuning, method ensembling, and the target biobank on PGS performance. We found that no single method consistently outperformed all others. PGS effect sizes were more variable between biobanks than between methods within biobanks when methods were well tuned. Differences between methods were largest for the two investigated autoimmune diseases, seropositive rheumatoid arthritis and type 1 diabetes. For most methods, cross-validation was more reliable for tuning hyperparameters than automatic tuning (without the use of target data). For a given target phenotype, elastic net models combining PGS across methods (ensemble PGS) tuned in the UK Biobank provided consistent, high, and cross-biobank transferable performance, increasing PGS effect sizes (ß coefficients) by a median of 5.0% relative to LDpred2 and MegaPRS (the two best-performing single methods when tuned with cross-validation). Our interactively browsable online-results and open-source workflow prspipe provide a rich resource and reference for the analysis of polygenic scoring methods across biobanks.
Assuntos
Bancos de Espécimes Biológicos , Estudo de Associação Genômica Ampla , Herança Multifatorial , Humanos , Herança Multifatorial/genética , Fenótipo , Diabetes Mellitus Tipo 1/genética , Polimorfismo de Nucleotídeo Único , Aprendizado de MáquinaRESUMO
BACKGROUND: Digitalization with minimal human resources could support self-management among women with gestational diabetes and improve maternal and neonatal outcomes. OBJECTIVE: This study aimed to investigate if a periodic mobile application (eMOM) with wearable sensors improves maternal and neonatal outcomes among women with diet-controlled gestational diabetes without additional guidance from healthcare personnel. STUDY DESIGN: Women with gestational diabetes were randomly assigned in a 1:1 ratio at 24 to 28 weeks' gestation to the intervention or the control arm. The intervention arm received standard care in combination with use of the periodic eMOM, whereas the control arm received only standard care. The intervention arm used eMOM with a continuous glucose monitor, an activity tracker, and a food diary 1 week/month until delivery. The primary outcome was the change in fasting plasma glucose from baseline to 35 to 37 weeks' gestation. Secondary outcomes included capillary glucose, weight gain, nutrition, physical activity, pregnancy complications, and neonatal outcomes, such as macrosomia. RESULTS: In total, 148 women (76 in the intervention arm, 72 in the control arm; average age, 34.1±4.0 years; body mass index, 27.1±5.0 kg/m2) were randomized. The intervention arm showed a lower mean change in fasting plasma glucose than the control arm (difference, -0.15 mmol/L vs -2.7 mg/mL; P=.022) and lower capillary fasting glucose levels (difference, -0.04 mmol/L vs -0.7 mg/mL; P=.002). The intervention arm also increased their intake of vegetables (difference, 11.8 g/MJ; P=.043), decreased their sedentary behavior (difference, -27.3 min/d; P=.043), and increased light physical activity (difference, 22.8 min/d; P=.009) when compared with the control arm. In addition, gestational weight gain was lower (difference, -1.3 kg; P=.015), and there were less newborns with macrosomia in the intervention arm (difference, -13.1 %; P=.036). Adherence to eMOM was high (daily use >90%), and the usage correlated with lower maternal fasting (P=.0006) and postprandial glucose levels (P=.017), weight gain (P=.028), intake of energy (P=.021) and carbohydrates (P=.003), and longer duration of the daily physical activity (P=.0006). There were no significant between-arm differences in terms of pregnancy complications. CONCLUSION: Self-tracking of lifestyle factors and glucose levels without additional guidance improves self-management and the treatment of gestational diabetes, which also benefits newborns. The results of this study support the use of digital self-management and education tools in maternity care.
RESUMO
We experiment with recent ensemble machine learning methods in estimating healthcare costs, utilizing Finnish data containing rich individual-level information on healthcare costs, socioeconomic status and diagnostic data from multiple registries. Our data are a random 10% sample (553,675 observations) from the Finnish population in 2017. Using annual healthcare cost in 2017 as a response variable, we compare the performance of Random forest, Gradient Boosting Machine (GBM) and eXtreme Gradient Boosting (XGBoost) to linear regression. As machine learning methods are often seen as unsuitable in risk adjustment applications because of their relative opaqueness, we also introduce visualizations from the machine learning literature to help interpret the contribution of individual variables to the prediction. Our results show that ensemble machine learning methods can improve predictive performance, with all of them significantly outperforming linear regression, and that a certain level of interpretation can be provided for them. We also find individual-level socioeconomic variables to improve prediction accuracy and that their effect is larger for machine learning methods. However, we find that the predictions used for funding allocations are sensitive to model selection, highlighting the need for comprehensive robustness testing when estimating risk adjustment models used in applications.
RESUMO
BACKGROUND: Gestational diabetes mellitus (GDM) is an increasing health risk for pregnant women as well as their children. Telehealth interventions targeted at the management of GDM have been shown to be effective, but they still require health care professionals for providing guidance and feedback. Feedback from wearable sensors has been suggested to support the self-management of GDM, but it is unknown how self-tracking should be designed in clinical care. OBJECTIVE: This study aimed to investigate how to support the self-management of GDM with self-tracking of continuous blood glucose and lifestyle factors without help from health care personnel. We examined comprehensive self-tracking from self-discovery (ie, learning associations between glucose levels and lifestyle) and user experience perspectives. METHODS: We conducted a mixed methods study where women with GDM (N=10) used a continuous glucose monitor (CGM; Medtronic Guardian) and 3 physical activity sensors: activity bracelet (Garmin Vivosmart 3), hip-worn sensor (UKK Exsed), and electrocardiography sensor (Firstbeat 2) for a week. We collected data from the sensors, and after use, participants took part in semistructured interviews about the wearable sensors. Acceptability of the wearable sensors was evaluated with the Unified Theory of Acceptance and Use of Technology (UTAUT) questionnaire. Moreover, maternal nutrition data were collected with a 3-day food diary, and self-reported physical activity data were collected with a logbook. RESULTS: We found that the CGM was the most useful sensor for the self-discovery process, especially when learning associations between glucose and nutrition intake. We identified new challenges for using data from the CGM and physical activity sensors in supporting self-discovery in GDM. These challenges included (1) dispersion of glucose and physical activity data in separate applications, (2) absence of important trackable features like amount of light physical activity and physical activities other than walking, (3) discrepancy in the data between different wearable physical activity sensors and between CGMs and capillary glucose meters, and (4) discrepancy in perceived and measured quantification of physical activity. We found the body placement of sensors to be a key factor in measurement quality and preference, and ultimately a challenge for collecting data. For example, a wrist-worn sensor was used for longer compared with a hip-worn sensor. In general, there was a high acceptance for wearable sensors. CONCLUSIONS: A mobile app that combines glucose, nutrition, and physical activity data in a single view is needed to support self-discovery. The design should support tracking features that are important for women with GDM (such as light physical activity), and data for each feature should originate from a single sensor to avoid discrepancy and redundancy. Future work with a larger sample should involve evaluation of the effects of such a mobile app on clinical outcomes. TRIAL REGISTRATION: Clinicaltrials.gov NCT03941652; https://clinicaltrials.gov/study/NCT03941652.
RESUMO
Methicillin-resistant Staphylococcus aureus (MRSA) is a major cause of morbidity and mortality. Colonization by MRSA increases the risk of infection and transmission, underscoring the importance of decolonization efforts. However, success of these decolonization protocols varies, raising the possibility that some MRSA strains may be more persistent than others. Here, we studied how the persistence of MRSA colonization correlates with genomic presence of antibiotic resistance genes. Our analysis using a Bayesian mixed effects survival model found that genetic determinants of high-level resistance to mupirocin was strongly associated with failure of the decolonization protocol. However, we did not see a similar effect with genetic resistance to chlorhexidine or other antibiotics. Including strain-specific random effects improved the predictive performance, indicating that some strain characteristics other than resistance also contributed to persistence. Study subject-specific random effects did not improve the model. Our results highlight the need to consider the properties of the colonizing MRSA strain when deciding which treatments to include in the decolonization protocol.
Assuntos
Anti-Infecciosos Locais , Staphylococcus aureus Resistente à Meticilina , Infecções Estafilocócicas , Humanos , Teorema de Bayes , Infecções Estafilocócicas/tratamento farmacológico , Portador Sadio , Antibacterianos/farmacologia , Resistência Microbiana a MedicamentosRESUMO
Cross-modality image synthesis is an active research topic with multiple medical clinically relevant applications. Recently, methods allowing training with paired but misaligned data have started to emerge. However, no robust and well-performing methods applicable to a wide range of real world data sets exist. In this work, we propose a generic solution to the problem of cross-modality image synthesis with paired but non-aligned data by introducing new deformation equivariance encouraging loss functions. The method consists of joint training of an image synthesis network together with separate registration networks and allows adversarial training conditioned on the input even with misaligned data. The work lowers the bar for new clinical applications by allowing effortless training of cross-modality image synthesis networks for more difficult data sets.
RESUMO
MOTIVATION: Existing methods for simulating synthetic genotype and phenotype datasets have limited scalability, constraining their usability for large-scale analyses. Moreover, a systematic approach for evaluating synthetic data quality and a benchmark synthetic dataset for developing and evaluating methods for polygenic risk scores are lacking. RESULTS: We present HAPNEST, a novel approach for efficiently generating diverse individual-level genotypic and phenotypic data. In comparison to alternative methods, HAPNEST shows faster computational speed and a lower degree of relatedness with reference panels, while generating datasets that preserve key statistical properties of real data. These desirable synthetic data properties enabled us to generate 6.8 million common variants and nine phenotypes with varying degrees of heritability and polygenicity across 1 million individuals. We demonstrate how HAPNEST can facilitate biobank-scale analyses through the comparison of seven methods to generate polygenic risk scoring across multiple ancestry groups and different genetic architectures. AVAILABILITY AND IMPLEMENTATION: A synthetic dataset of 1 008 000 individuals and nine traits for 6.8 million common variants is available at https://www.ebi.ac.uk/biostudies/studies/S-BSST936. The HAPNEST software for generating synthetic datasets is available as Docker/Singularity containers and open source Julia and C code at https://github.com/intervene-EU-H2020/synthetic_data.
Assuntos
Benchmarking , Confiabilidade dos Dados , Humanos , Genótipo , Fenótipo , Herança MultifatorialRESUMO
Subspace distance is an invaluable tool exploited in a wide range of feature selection methods. The power of subspace distance is that it can identify a representative subspace, including a group of features that can efficiently approximate the space of original features. On the other hand, employing intrinsic statistical information of data can play a significant role in a feature selection process. Nevertheless, most of the existing feature selection methods founded on the subspace distance are limited in properly fulfilling this objective. To pursue this void, we propose a framework that takes a subspace distance into account which is called "Variance-Covariance subspace distance". The approach gains advantages from the correlation of information included in the features of data, thus determines all the feature subsets whose corresponding Variance-Covariance matrix has the minimum norm property. Consequently, a novel, yet efficient unsupervised feature selection framework is introduced based on the Variance-Covariance distance to handle both the dimensionality reduction and subspace learning tasks. The proposed framework has the ability to exclude those features that have the least variance from the original feature set. Moreover, an efficient update algorithm is provided along with its associated convergence analysis to solve the optimization side of the proposed approach. An extensive number of experiments on nine benchmark datasets are also conducted to assess the performance of our method from which the results demonstrate its superiority over a variety of state-of-the-art unsupervised feature selection methods. The source code is available at https://github.com/SaeedKarami/VCSDFS.
Assuntos
Algoritmos , Reconhecimento Automatizado de Padrão , Reconhecimento Automatizado de Padrão/métodos , Aprendizagem , Software , BenchmarkingRESUMO
Background: Weight loss effectively reduces cardiometabolic health risks among people with overweight and obesity, but inter-individual variability in weight loss maintenance is large. Here we studied whether baseline gene expression in subcutaneous adipose tissue predicts diet-induced weight loss success. Methods: Within the 8-month multicenter dietary intervention study DiOGenes, we classified a low weight-losers (low-WL) group and a high-WL group based on median weight loss percentage (9.9%) from 281 individuals. Using RNA sequencing, we identified the significantly differentially expressed genes between high-WL and low-WL at baseline and their enriched pathways. We used this information together with support vector machines with linear kernel to build classifier models that predict the weight loss classes. Results: Prediction models based on a selection of genes that are associated with the discovered pathways 'lipid metabolism' (max AUC = 0.74, 95% CI [0.62-0.86]) and 'response to virus' (max AUC = 0.72, 95% CI [0.61-0.83]) predicted the weight-loss classes high-WL/low-WL significantly better than models based on randomly selected genes (P < 0.01). The performance of the models based on 'response to virus' genes is highly dependent on those genes that are also associated with lipid metabolism. Incorporation of baseline clinical factors into these models did not noticeably enhance the model performance in most of the runs. This study demonstrates that baseline adipose tissue gene expression data, together with supervised machine learning, facilitates the characterization of the determinants of successful weight loss.
Assuntos
Dieta Redutora , Obesidade , Humanos , Obesidade/genética , Gordura Subcutânea/metabolismo , Redução de Peso/genética , Expressão Gênica/genética , LipídeosRESUMO
Researchers aim to develop polygenic risk scores as a tool to prevent and more effectively treat serious diseases, disorders and conditions such as breast cancer, type 2 diabetes mellitus and coronary heart disease. Recently, machine learning techniques, in particular deep neural networks, have been increasingly developed to create polygenic risk scores using electronic health records as well as genomic and other health data. While the use of artificial intelligence for polygenic risk scores may enable greater accuracy, performance and prediction, it also presents a range of increasingly complex ethical challenges. The ethical and social issues of many polygenic risk score applications in medicine have been widely discussed. However, in the literature and in practice, the ethical implications of their confluence with the use of artificial intelligence have not yet been sufficiently considered. Based on a comprehensive review of the existing literature, we argue that this stands in need of urgent consideration for research and subsequent translation into the clinical setting. Considering the many ethical layers involved, we will first give a brief overview of the development of artificial intelligence-driven polygenic risk scores, associated ethical and social implications, challenges in artificial intelligence ethics, and finally, explore potential complexities of polygenic risk scores driven by artificial intelligence. We point out emerging complexity regarding fairness, challenges in building trust, explaining and understanding artificial intelligence and polygenic risk scores as well as regulatory uncertainties and further challenges. We strongly advocate taking a proactive approach to embedding ethics in research and implementation processes for polygenic risk scores driven by artificial intelligence.
RESUMO
The extent to which genetic interactions affect observed phenotypes is generally unknown because current interaction detection approaches only consider simple interactions between top SNPs of genes. We introduce an open-source framework for increasing the power of interaction detection by considering all SNPs within a selected set of genes and complex interactions between them, beyond only the currently considered multiplicative relationships. In brief, the relation between SNPs and a phenotype is captured by a neural network, and the interactions are quantified by Shapley scores between hidden nodes, which are gene representations that optimally combine information from the corresponding SNPs. Additionally, we design a permutation procedure tailored for neural networks to assess the significance of interactions, which outperformed existing alternatives on simulated datasets with complex interactions, and in a cholesterol study on the UK Biobank it detected nine interactions which replicated on an independent FINRISK dataset.
Assuntos
Aprendizado Profundo , Epistasia Genética , Redes Neurais de Computação , Polimorfismo de Nucleotídeo Único , FenótipoRESUMO
INTRODUCTION: Gestational diabetes (GDM) causes various adverse short-term and long-term consequences for the mother and child, and its incidence is increasing globally. So far, the most promising digital health interventions for GDM management have involved healthcare professionals to provide guidance and feedback. The principal aim of this study is to evaluate the effects of comprehensive and real-time self-tracking with eMOM GDM mobile application (app) on glucose levels in women with GDM, and more broadly, on different other maternal and neonatal outcomes. METHODS AND ANALYSIS: This randomised controlled trial is carried out in Helsinki metropolitan area. We randomise 200 pregnant women with GDM into the intervention and the control group at gestational week (GW) 24-28 (baseline, BL). The intervention group receives standard antenatal care and the eMOM GDM app, while the control group will receive only standard care. Participants in the intervention group use the eMOM GDM app with continuous glucose metre (CGM) and activity bracelet for 1 week every month until delivery and an electronic 3-day food record every month until delivery. The follow-up visit after intervention takes place 3 months post partum for both groups. Data are collected by laboratory blood tests, clinical measurements, capillary glucose measures, wearable sensors, air displacement plethysmography and digital questionnaires. The primary outcome is fasting plasma glucose change from BL to GW 35-37. Secondary outcomes include, for example, self-tracked capillary fasting and postprandial glucose measures, change in gestational weight gain, change in nutrition quality, change in physical activity, medication use due to GDM, birth weight and fat percentage of the child. ETHICS AND DISSEMINATION: The study has been approved by Ethics Committee of the Helsinki and Uusimaa Hospital District. The results will be presented in peer-reviewed journals and at conferences. TRIAL REGISTRATION NUMBER: NCT04714762.
Assuntos
Diabetes Gestacional , Aplicativos Móveis , Recém-Nascido , Criança , Feminino , Gravidez , Humanos , Diabetes Gestacional/epidemiologia , Glicemia , Estilo de Vida , Peso ao Nascer , Ensaios Clínicos Controlados Aleatórios como AssuntoRESUMO
Methicillin-resistant Staphylococcus aureus (MRSA) can colonize multiple body sites, and carriage is a risk factor for infection. Successful decolonization protocols reduce disease incidence; however, multiple protocols exist, comprising diverse therapies targeting multiple body sites, and the optimal protocol is unclear. Standard methods cannot infer the impact of site-specific components on successful decolonization. Here, we formulate a Bayesian coupled hidden Markov model, which estimates interactions between body sites, quantifies the contribution of each therapy to successful decolonization, and enables predictions of the efficacy of therapy combinations. We applied the model to longitudinal data from a randomized controlled trial (RCT) of an MRSA decolonization protocol consisting of chlorhexidine body and mouthwash and nasal mupirocin. Our findings (i) confirmed nares as a central hub for MRSA colonization and nasal mupirocin as the most crucial therapy and (ii) demonstrated all components contributed significantly to the efficacy of the protocol and the protocol reduced self-inoculation. Finally, we assessed the impact of hypothetical therapy improvements in silico and found that enhancing MRSA clearance at the skin would yield the largest gains. This study demonstrates the use of advanced modelling to go beyond what is typically achieved by RCTs, enabling evidence-based decision-making to streamline clinical protocols.
Assuntos
Anti-Infecciosos Locais , Staphylococcus aureus Resistente à Meticilina , Infecções Estafilocócicas , Antibacterianos/farmacologia , Antibacterianos/uso terapêutico , Anti-Infecciosos Locais/uso terapêutico , Portador Sadio/tratamento farmacológico , Humanos , Mupirocina/farmacologia , Mupirocina/uso terapêutico , Infecções Estafilocócicas/tratamento farmacológico , Infecções Estafilocócicas/epidemiologiaRESUMO
Human knowledge provides a formal understanding of the world. Knowledge graphs that represent structural relations between entities have become an increasingly popular research direction toward cognition and human-level intelligence. In this survey, we provide a comprehensive review of the knowledge graph covering overall research topics about: 1) knowledge graph representation learning; 2) knowledge acquisition and completion; 3) temporal knowledge graph; and 4) knowledge-aware applications and summarize recent breakthroughs and perspective directions to facilitate future research. We propose a full-view categorization and new taxonomies on these topics. Knowledge graph embedding is organized from four aspects of representation space, scoring function, encoding models, and auxiliary information. For knowledge acquisition, especially knowledge graph completion, embedding methods, path inference, and logical rule reasoning are reviewed. We further explore several emerging topics, including metarelational learning, commonsense reasoning, and temporal knowledge graphs. To facilitate future research on knowledge graphs, we also provide a curated collection of data sets and open-source libraries on different tasks. In the end, we have a thorough outlook on several promising research directions.
RESUMO
Legionella pneumophila is the most common cause of the severe respiratory infection known as Legionnaires' disease. However, the microorganism is typically a symbiont of free-living amoeba, and our understanding of the bacterial factors that determine human pathogenicity is limited. Here we carried out a population genomic study of 902 L. pneumophila isolates from human clinical and environmental samples to examine their genetic diversity, global distribution and the basis for human pathogenicity. We find that the capacity for human disease is representative of the breadth of species diversity although some clones are more commonly associated with clinical infections. We identified a single gene (lag-1) to be most strongly associated with clinical isolates. lag-1, which encodes an O-acetyltransferase for lipopolysaccharide modification, has been distributed horizontally across all major phylogenetic clades of L. pneumophila by frequent recent recombination events. The gene confers resistance to complement-mediated killing in human serum by inhibiting deposition of classical pathway molecules on the bacterial surface. Furthermore, acquisition of lag-1 inhibits complement-dependent phagocytosis by human neutrophils, and promoted survival in a mouse model of pulmonary legionellosis. Thus, our results reveal L. pneumophila genetic traits linked to disease and provide a molecular basis for resistance to complement-mediated killing.
Assuntos
Proteínas do Sistema Complemento/imunologia , Legionella pneumophila/genética , Doença dos Legionários/imunologia , Doença dos Legionários/microbiologia , Acetiltransferases/genética , Acetiltransferases/imunologia , Animais , Proteínas de Bactérias/genética , Proteínas de Bactérias/imunologia , Feminino , Genoma Bacteriano , Humanos , Legionella pneumophila/classificação , Legionella pneumophila/imunologia , Legionella pneumophila/isolamento & purificação , Camundongos , Camundongos Endogâmicos C57BL , Neutrófilos/imunologia , FilogeniaRESUMO
Identifying crop loss at field parcel scale using satellite images is challenging: first, crop loss is caused by many factors during the growing season; second, reliable reference data about crop loss are lacking; third, there are many ways to define crop loss. This study investigates the feasibility of using satellite images to train machine learning (ML) models to classify agricultural field parcels into those with and without crop loss. The reference data for this study was provided by Finnish Food Authority (FFA) containing crop loss information of approximately 1.4 million field parcels in Finland covering about 3.5 million ha from 2000 to 2015. This reference data was combined with Normalised Difference Vegetation Index (NDVI) derived from Landsat 7 images, in which more than 80% of the possible data are missing. Despite the hard problem with extremely noisy data, among the four ML models we tested, random forest (with mean imputation and missing value indicators) achieved the average AUC (area under the ROC curve) of 0.688±0.059 over all 16 years with the range [0.602, 0.795] in identifying new crop-loss fields based on reference fields of the same year. To our knowledge, this is one of the first large scale benchmark study of using machine learning for crop loss classification at field parcel scale. The classification setting and trained models have numerous potential applications, for example, allowing government agencies or insurance companies to verify crop-loss claims by farmers and realise efficient agricultural monitoring.
Assuntos
Produtos Agrícolas/crescimento & desenvolvimento , Aprendizado de Máquina , Imagens de Satélites , Estações do Ano , FinlândiaRESUMO
Unsupervised pretraining is an integral part of many natural language processing systems, and transfer learning with language models has achieved remarkable results in downstream tasks. In the clinical application of medical code assignment, diagnosis and procedure codes are inferred from lengthy clinical notes such as hospital discharge summaries. However, it is not clear if pretrained models are useful for medical code prediction without further architecture engineering. This paper conducts a comprehensive quantitative analysis of various contextualized language models' performances, pretrained in different domains, for medical code assignment from clinical notes. We propose a hierarchical fine-tuning architecture to capture interactions between distant words and adopt label-wise attention to exploit label information. Contrary to current trends, we demonstrate that a carefully trained classical CNN outperforms attention-based models on a MIMIC-III subset with frequent codes. Our empirical findings suggest directions for building robust medical code assignment models.
Assuntos
Idioma , Processamento de Linguagem NaturalRESUMO
OBJECTIVES: Our aim was to investigate in a real-life setting the use of machine learning for modelling the postprandial glucose concentrations in morbidly obese patients undergoing Roux-en-Y gastric bypass (RYGB) or one-anastomosis gastric bypass (OAGB). METHODS: As part of the prospective randomized open-label trial (RYSA), data from obese (BMI ≥35 kg/m2) non-diabetic adult participants were included. Glucose concentrations, measured with FreeStyle Libre, were recorded over 14 preoperative and 14 postoperative days. During these periods, 3-day food intake was self-reported. A machine learning model was applied to estimate glycaemic responses to the reported carbohydrate intakes before and after the bariatric surgeries. RESULTS: Altogether, 10 participants underwent RYGB and 7 participants OAGB surgeries. The glucose concentrations and carbohydrate intakes were reduced postoperatively in both groups. The relative time spent in hypoglycaemia increased regardless of the operation (RYGB, from 9.2 to 28.2%; OAGB, from 1.8 to 37.7%). Postoperatively, we observed an increase in the height of the fitted response curve and a reduction in its width, suggesting that the same amount of carbohydrates caused a larger increase in the postprandial glucose response and that the clearance of the meal-derived blood glucose was faster, with no clinically meaningful differences between the surgeries. CONCLUSIONS: A detailed analysis of the glycaemic responses using food diaries has previously been difficult because of the noisy meal data. The utilized machine learning model resolved this by modelling the uncertainty in meal times. Such an approach is likely also applicable in other applications involving dietary data. A marked reduction in overall glycaemia, increase in postprandial glucose response, and rapid glucose clearance from the circulation immediately after surgery are evident after both RYGB and OAGB. Whether nondiabetic individuals would benefit from monitoring the post-surgery hypoglycaemias and the potential to prevent them by dietary means should be investigated.KEY MESSAGESThe use of a novel machine learning model was applicable for combining patient-reported data and time-series data in this clinical study.Marked increase in postprandial glucose concentrations and rapid glucose clearance were observed after both Roux-en-Y gastric bypass and one-anastomosis gastric bypass surgeries.Whether nondiabetic individuals would benefit from monitoring the post-surgery hypoglycaemias and the potential to prevent them by dietary means should be investigated.
Assuntos
Anastomose em-Y de Roux/estatística & dados numéricos , Glicemia , Carboidratos da Dieta/administração & dosagem , Gastrectomia/estatística & dados numéricos , Derivação Gástrica/estatística & dados numéricos , Obesidade Mórbida/cirurgia , Adulto , Simulação por Computador , Feminino , Humanos , Masculino , Pessoa de Meia-Idade , Estudos Prospectivos , AutorrelatoRESUMO
Estimating the impact of a treatment on a given response is needed in many biomedical applications. However, methodology is lacking for the case when the response is a continuous temporal curve, treatment covariates suffer extensively from measurement error, and even the exact timing of the treatments is unknown. We introduce a novel method for this challenging scenario. We model personalized treatment-response curves as a combination of parametric response functions, hierarchically sharing information across individuals, and a sparse Gaussian process for the baseline trend. Importantly, our model accounts for errors not only in treatment covariates, but also in treatment timings, a problem arising in practice for example when data on treatments are based on user self-reporting. We validate our model with simulated and real patient data, and show that in a challenging application of estimating the impact of diet on continuous blood glucose measurements, accounting for measurement error significantly improves estimation and prediction accuracy.