Search

Applying deep learning, especially time series neural networks, to predict stock price, has become one of the important applications in quantitative finance. Recently, some GAN-based stock prediction models are proposed, where LSTM or GRU is used as the generator. However, these generators lack the function of feature extraction, and the prediction accuracies are slightly low. Meanwhile, these models choose some simple volume-price factors (such as OCHLV and OCHLVC) as inputs, without considering the impact of other factors on stock prices. In order to solve these problems, a stock prediction method based on multiple factors and GAN-TrellisNet is proposed. Instead of "OCHLV" or "OCHLVC," a multi-factor strategy with "alpha158+OCHLVC" is introduced to enrich the stock data of inputs. The proposed generative adversarial network (GAN) is a combination of two neural networks which are TrellisNet as generative model and convolutional neural network (CNN) as discriminative model for adversarial training to forecast the stock market. TrellisNet, which integrates the feature extraction capabilities of CNN and the temporal processing capabilities of recurrent neural network (RNN), will generate new predicted results based on historical data, and then CNN will distinguish between predicted results and real stock prices. In order to demonstrate the performance of our method, we selected the decade data of different stocks from four markets (A-shares, U.S. stocks, U.K. stocks and Hong Kong stocks) as dataset and conducted two groups of comparative experiments. Compared with the state-of-the-art methods based on GAN, our method has better performance in terms of MSE, MAE, RMSE and MAPE. In addition, the multi-factor strategy with "alpha158+OCHLVC" is more effective than the original strategy with OCHLVC factors. [ABSTRACT FROM AUTHOR]

Previous research indicates that discipline identification — the level of integration of one's discipline of study into the self-concept — is associated with deep approaches to learning and academic self-efficacy. However, it is not known whether these relationships would hold in the context of learning during the COVID-19 pandemic. The present study investigates this, as well as exploring whether belonging, identity incompatibility, and virtual and in-person time spent with fellow students and with faculty act as potential antecedents of discipline identification. A cross-sectional online survey with n = 385 students in the UK was conducted during the spring term (February–May) of the 2019–2020 academic year, when UK universities were under COVID-19 restrictions. Of those students, n = 109 completed a follow-up survey during the autumn term (September–December) of the 2020–2021 academic year. Due to pandemic-related changes, the two waves of data were analysed separately using path analysis. During the spring term, identity incompatibility and discipline identification were significantly associated with both deep approaches to learning and academic self-efficacy. There were also significant indirect effects of belonging on both deep approaches to learning and academic self-efficacy via discipline identification. During the autumn term, discipline identification was again significantly associated with both academic outcomes. Identity incompatibility was significantly associated with academic self-efficacy. There were also indirect effects of identity incompatibility and belonging on deep approaches to learning via discipline identification. These results add to the literature on the importance of discipline identification and associated social-psychological processes in predicting important higher education outcomes during the COVID-19 pandemic. [ABSTRACT FROM AUTHOR]

In recent years, the utility of polygenic risk scores (PRS) in forecasting disease susceptibility from genome-wide association studies (GWAS) results has been widely recognised. Yet, these models face limitations due to overfitting and the potential overestimation of effect sizes in correlated variants. To surmount these obstacles, we devised the Stacked Neural Network Polygenic Risk Score (SNPRS). This novel approach synthesises outputs from multiple neural network models, each calibrated using genetic variants chosen based on diverse p-value thresholds. By doing so, SNPRS captures a broader array of genetic variants, enabling a more nuanced interpretation of the combined effects of these variants. We assessed the efficacy of SNPRS using the UK Biobank data, focusing on the genetic risks associated with breast and prostate cancers, as well as quantitative traits like height and BMI. We also extended our analysis to the Korea Genome and Epidemiology Study (KoGES) dataset. Impressively, our results indicate that SNPRS surpasses traditional PRS models and an isolated deep neural network in terms of accuracy, highlighting its promise in refining the efficacy and relevance of PRS in genetic studies. [ABSTRACT FROM AUTHOR]

Increased left atrial volume and decreased left atrial function have long been associated with atrial fibrillation. The availability of large-scale cardiac magnetic resonance imaging data paired with genetic data provides a unique opportunity to assess the genetic contributions to left atrial structure and function, and understand their relationship with risk for atrial fibrillation. Here, we use deep learning and surface reconstruction models to measure left atrial minimum volume, maximum volume, stroke volume, and emptying fraction in 40,558 UK Biobank participants. In a genome-wide association study of 35,049 participants without pre-existing cardiovascular disease, we identify 20 common genetic loci associated with left atrial structure and function. We find that polygenic contributions to increased left atrial volume are associated with atrial fibrillation and its downstream consequences, including stroke. Through Mendelian randomization, we find evidence supporting a causal role for left atrial enlargement and dysfunction on atrial fibrillation risk. In this study, a deep learning-based model of left atrial size in UK Biobank enabled genome-wide association studies in 35,049 healthy participants. Several lines of evidence, including the PITX2 locus, linked left atrial dysfunction to atrial fibrillation risk. [ABSTRACT FROM AUTHOR]

Understanding the genetic architecture of brain structure is challenging, partly due to difficulties in designing robust, non-biased descriptors of brain morphology. Until recently, brain measures for genome-wide association studies (GWAS) consisted of traditionally expert-defined or software-derived image-derived phenotypes (IDPs) that are often based on theoretical preconceptions or computed from limited amounts of data. Here, we present an approach to derive brain imaging phenotypes using unsupervised deep representation learning. We train a 3-D convolutional autoencoder model with reconstruction loss on 6130 UK Biobank (UKBB) participants' T1 or T2-FLAIR (T2) brain MRIs to create a 128-dimensional representation known as Unsupervised Deep learning derived Imaging Phenotypes (UDIPs). GWAS of these UDIPs in held-out UKBB subjects (n = 22,880 discovery and n = 12,359/11,265 replication cohorts for T1/T2) identified 9457 significant SNPs organized into 97 independent genetic loci of which 60 loci were replicated. Twenty-six loci were not reported in earlier T1 and T2 IDP-based UK Biobank GWAS. We developed a perturbation-based decoder interpretation approach to show that these loci are associated with UDIPs mapped to multiple relevant brain regions. Our results established unsupervised deep learning can derive robust, unbiased, heritable, and interpretable brain imaging phenotypes. A study utilizing unsupervised deep learning to generate interpretable brain imaging phenotypes from brain T1 and T2-FLAIR MRI identified 97 genetic loci enhancing understanding of brain structure genetics. [ABSTRACT FROM AUTHOR]

Objectives: There is a need for CT pulmonary angiography (CTPA) lung segmentation models. Clinical translation requires radiological evaluation of model outputs, understanding of limitations, and identification of failure points. This multicentre study aims to develop an accurate CTPA lung segmentation model, with evaluation of outputs in two diverse patient cohorts with pulmonary hypertension (PH) and interstitial lung disease (ILD). Methods: This retrospective study develops an nnU-Net-based segmentation model using data from two specialist centres (UK and USA). Model was trained (n = 37), tested (n = 12), and clinically evaluated (n = 176) on a diverse 'real-world' cohort of 225 PH patients with volumetric CTPAs. Dice score coefficient (DSC) and normalised surface distance (NSD) were used for testing. Clinical evaluation of outputs was performed by two radiologists who assessed clinical significance of errors. External validation was performed on heterogenous contrast and non-contrast scans from 28 ILD patients. Results: A total of 225 PH and 28 ILD patients with diverse demographic and clinical characteristics were evaluated. Mean accuracy, DSC, and NSD scores were 0.998 (95% CI 0.9976, 0.9989), 0.990 (0.9840, 0.9962), and 0.983 (0.9686, 0.9972) respectively. There were no segmentation failures. On radiological review, 82% and 71% of internal and external cases respectively had no errors. Eighteen percent and 25% respectively had clinically insignificant errors. Peripheral atelectasis and consolidation were common causes for suboptimal segmentation. One external case (0.5%) with patulous oesophagus had a clinically significant error. Conclusion: State-of-the-art CTPA lung segmentation model provides accurate outputs with minimal clinical errors on evaluation across two diverse cohorts with PH and ILD. Clinical relevance: Clinical translation of artificial intelligence models requires radiological review and understanding of model limitations. This study develops an externally validated state-of-the-art model with robust radiological review. Intended clinical use is in techniques such as lung volume or parenchymal disease quantification. Key Points: • Accurate, externally validated CT pulmonary angiography (CTPA) lung segmentation model tested in two large heterogeneous clinical cohorts (pulmonary hypertension and interstitial lung disease). • No segmentation failures and robust review of model outputs by radiologists found 1 (0.5%) clinically significant segmentation error. • Intended clinical use of this model is a necessary step in techniques such as lung volume, parenchymal disease quantification, or pulmonary vessel analysis. [ABSTRACT FROM AUTHOR]

Across the globe, governments are developing policies and strategies to reduce carbon emissions to address climate change. Monitoring the impact of governments' carbon reduction policies can significantly enhance our ability to combat climate change and meet emissions reduction targets. One promising area in this regard is the role of artificial intelligence (AI) in carbon reduction policy and strategy monitoring. While researchers have explored applications of AI on data from various sources, including sensors, satellites, and social media, to identify areas for carbon emissions reduction, AI applications in tracking the effect of governments' carbon reduction plans have been limited. This study presents an AI framework based on long short-term memory (LSTM) and statistical process control (SPC) for the monitoring of variations in carbon emissions, using UK annual CO2 emission (per capita) data, covering a period between 1750 and 2021. This paper used LSTM to develop a surrogate model for the UK's carbon emissions characteristics and behaviours. As observed in our experiments, LSTM has better predictive abilities than ARIMA, Exponential Smoothing and feedforward artificial neural networks (ANN) in predicting CO2 emissions on a yearly prediction horizon. Using the deviation of the recorded emission data from the surrogate process, the variations and trends in these behaviours are then analysed using SPC, specifically Shewhart individual/moving range control charts. The result shows several assignable variations between the mid-1990s and 2021, which correlate with some notable UK government commitments to lower carbon emissions within this period. The framework presented in this paper can help identify periods of significant deviations from a country's normal CO2 emissions, which can potentially result from the government's carbon reduction policies or activities that can alter the amount of CO2 emissions. [ABSTRACT FROM AUTHOR]

For efficient operation, solar power operators often require generation forecasts for multiple sites with varying data availability. Many proposed methods for forecasting solar irradiance / solar power production formulate the problem as a time-series, using current observations to generate forecasts. This necessitates a real-time data stream and enough historical observations at every location for these methods to be deployed. In this paper, we propose the use of Global methods to train generalised models. Using data from 20 locations distributed throughout the UK, we show that it is possible to learn models without access to data for all locations, enabling them to generate forecasts for unseen locations. We show a single Global model trained on multiple locations can produce more consistent and accurate results across locations. Furthermore, by leveraging weather observations and measurements from other locations we show it is possible to create models capable of accurately forecasting irradiance at locations without any real-time data. We apply our approaches to both classical and state-of-the-art Machine Learning methods, including a Transformer architecture. We compare models using satellite imagery or point observations (temperature, pressure, etc.) as weather data. These methods could facilitate planning and optimisation for both newly deployed solar farms and domestic installations from the moment they come online. [ABSTRACT FROM AUTHOR]

Researchers and engineers employ machine learning (ML) tools to detect pipe bursts and prevent significant non-revenue water losses in water distribution systems (WDS). Nonetheless, many approaches developed so far consider a fixed number of sensors, which requires the ML model redevelopment and collection of sufficient data with the new sensor configuration for training. To overcome these issues, this study presents a novel approach based on Long Short-Term Memory neural networks (NNs) that leverages transfer learning to manage a varying number of sensors and retain good detection performance with limited training data. The proposed detection model first learns to reproduce the normal behavior of the system on a dataset obtained in burst-free conditions. The training process involves predicting flow and pressure one-time step ahead using historical data and time-related features as inputs. During testing, a post-prediction step flags potential bursts based on the comparison between the observations and model predictions using a time-varied error threshold. When adding new sensors, we implement transfer learning by replicating the weights of existing channels and then fine-tune the augmented NN. We evaluate the robustness of the methodology on simulated fire hydrant bursts and real-bursts in 10 district metered areas (DMAs) of the UK. For real bursts, we perform a sensitivity analysis to understand the impact of data resolution and error threshold on burst detection performance. The results obtained demonstrate that this ML-based methodology can achieve Precision of up to 98.1% in real-life settings and can identify bursts, even in data scarce conditions. [ABSTRACT FROM AUTHOR]

Objective: An increasing number of commercial deep learning computer-aided detection (DL-CAD) systems are available but their cost-saving potential is largely unknown. This study aimed to gain insight into appropriate pricing for DL-CAD in different reading modes to be cost-saving and to determine the potentially most cost-effective reading mode for lung cancer screening. Methods: In three representative settings, DL-CAD was evaluated as a concurrent, pre-screening, and second reader. Scoping review was performed to estimate radiologist reading time with and without DL-CAD. Hourly cost of radiologist time was collected for the USA (€196), UK (€127), and Poland (€45), and monetary equivalence of saved time was calculated. The minimum number of screening CTs to reach break-even was calculated for one-time investment of €51,616 for DL-CAD. Results: Mean reading time was 162 (95% CI: 111–212) seconds per case without DL-CAD, which decreased by 77 (95% CI: 47–107) and 104 (95% CI: 71–136) seconds for DL-CAD as concurrent and pre-screening reader, respectively, and increased by 33–41 s for DL-CAD as second reader. This translates into €1.0–4.3 per-case cost for concurrent reading and €0.8–5.7 for pre-screening reading in the USA, UK, and Poland. To achieve break-even with a one-time investment, the minimum number of CT scans was 12,300–53,600 for concurrent reader, and 9400–65,000 for pre-screening reader in the three countries. Conclusions: Given current pricing, DL-CAD must be priced substantially below €6 in a pay-per-case setting or used in a high-workload environment to reach break-even in lung cancer screening. DL-CAD as pre-screening reader shows the largest potential to be cost-saving. Critical relevance statement: Deep-learning computer-aided lung nodule detection (DL-CAD) software must be priced substantially below 6 euro in a pay-per-case setting or must be used in high-workload environments with one-time investment in order to achieve break-even. DL-CAD as a pre-screening reader has the greatest cost savings potential. Key points: • DL-CAD must be substantially below €6 in a pay-per-case setting to reach break-even. • DL-CAD must be used in a high-workload screening environment to achieve break-even. • DL-CAD as a pre-screening reader shows the largest potential to be cost-saving. [ABSTRACT FROM AUTHOR]

Despite the importance of preventing chronic kidney disease (CKD), predicting high-risk patients who require active intervention is challenging, especially in people with preserved kidney function. In this study, a predictive risk score for CKD (Reti-CKD score) was derived from a deep learning algorithm using retinal photographs. The performance of the Reti-CKD score was verified using two longitudinal cohorts of the UK Biobank and Korean Diabetic Cohort. Validation was done in people with preserved kidney function, excluding individuals with eGFR <90 mL/min/1.73 m2 or proteinuria at baseline. In the UK Biobank, 720/30,477 (2.4%) participants had CKD events during the 10.8-year follow-up period. In the Korean Diabetic Cohort, 206/5014 (4.1%) had CKD events during the 6.1-year follow-up period. When the validation cohorts were divided into quartiles of Reti-CKD score, the hazard ratios for CKD development were 3.68 (95% Confidence Interval [CI], 2.88–4.41) in the UK Biobank and 9.36 (5.26–16.67) in the Korean Diabetic Cohort in the highest quartile compared to the lowest. The Reti-CKD score, compared to eGFR based methods, showed a superior concordance index for predicting CKD incidence, with a delta of 0.020 (95% CI, 0.011–0.029) in the UK Biobank and 0.024 (95% CI, 0.002–0.046) in the Korean Diabetic Cohort. In people with preserved kidney function, the Reti-CKD score effectively stratifies future CKD risk with greater performance than conventional eGFR-based methods. [ABSTRACT FROM AUTHOR]

In recent years, the growing impact of climate change on surface water bodies has made the analysis and forecasting of streamflow rates essential for proper planning and management of water resources. This study proposes a novel ensemble (or hybrid) model, based on the combination of a Deep Learning algorithm, the Nonlinear AutoRegressive network with eXogenous inputs, and two Machine Learning algorithms, Multilayer Perceptron and Random Forest, for the short-term streamflow forecasting, considering precipitation as the only exogenous input and a forecast horizon up to 7 days. A large regional study was performed, considering 18 watercourses throughout the United Kingdom, characterized by different catchment areas and flow regimes. In particular, the predictions obtained with the ensemble Machine Learning-Deep Learning model were compared with the ones achieved with simpler models based on an ensemble of both Machine Learning algorithms and on the only Deep Learning algorithm. The hybrid Machine Learning-Deep Learning model outperformed the simpler models, with values of R2 above 0.9 for several watercourses, with the greatest discrepancies for small basins, where high and non-uniform rainfall throughout the year makes the streamflow rate forecasting a challenging task. Furthermore, the hybrid Machine Learning-Deep Learning model has been shown to be less affected by reductions in performance as the forecasting horizon increases compared to the simpler models, leading to reliable predictions even for 7-day forecasts. [ABSTRACT FROM AUTHOR]

Chemical-shift encoded MRI (CSE-MRI) is a widely used technique for the study of body composition and metabolic disorders, where derived fat and water signals enable the quantification of adipose tissue and muscle. The UK Biobank is acquiring whole-body Dixon MRI (a specific implementation of CSE-MRI) for over 100,000 participants. Current processing methods associated with large whole-body volumes are time intensive and prone to artifacts during fat-water separation performed by the scanner, making quantitative analysis challenging. The most common artifacts are fat-water swaps, where the labels are inverted at the voxel level. It is common for researchers to discard swapped data (generally around 10%), which is wasteful and may lead to unintended biases. Given the large number of whole-body Dixon MRI acquisitions in the UK Biobank, thousands of swaps are expected to be present in the fat and water volumes from image reconstruction performed on the scanner. If they go undetected, errors will propagate into processes such as organ segmentation, and dilute the results in population-based analyses. There is a clear need for a robust method to accurately separate fat and water volumes in big data collections like the UK Biobank. We formulate fat-water separation as a style transfer problem, where swap-free fat and water volumes are predicted from the acquired Dixon MRI data using a conditional generative adversarial network, and introduce a new loss function for the generator model. Our method is able to predict highly accurate fat and water volumes free from artifacts in the UK Biobank. We show that our model separates fat and water volumes using either single input (in-phase only) or dual input (in-phase and opposed-phase) data, with the latter producing superior results. Our proposed method enables faster and more accurate downstream analysis of body composition from Dixon MRI in population studies by eliminating the need for visual inspection or discarding data due to fat-water swaps. [ABSTRACT FROM AUTHOR]

Cardiorespiratory fitness is an established predictor of metabolic disease and mortality. Fitness is directly measured as maximal oxygen consumption (VO2max), or indirectly assessed using heart rate responses to standard exercise tests. However, such testing is costly and burdensome because it requires specialized equipment such as treadmills and oxygen masks, limiting its utility. Modern wearables capture dynamic real-world data which could improve fitness prediction. In this work, we design algorithms and models that convert raw wearable sensor data into cardiorespiratory fitness estimates. We validate these estimates' ability to capture fitness profiles in free-living conditions using the Fenland Study (N=11,059), along with its longitudinal cohort (N = 2675), and a third external cohort using the UK Biobank Validation Study (N = 181) who underwent maximal VO2max testing, the gold standard measurement of fitness. Our results show that the combination of wearables and other biomarkers as inputs to neural networks yields a strong correlation to ground truth in a holdout sample (r = 0.82, 95CI 0.80–0.83), outperforming other approaches and models and detects fitness change over time (e.g., after 7 years). We also show how the model's latent space can be used for fitness-aware patient subtyping paving the way to scalable interventions and personalized trial recruitment. These results demonstrate the value of wearables for fitness estimation that today can be measured only with laboratory tests. [ABSTRACT FROM AUTHOR]

Large epidemiological studies such as the UK Biobank (UKBB) or German National Cohort (NAKO) provide unprecedented health-related data of the general population aiming to better understand determinants of health and disease. As part of these studies, Magnetic Resonance Imaging (MRI) is performed in a subset of participants allowing for phenotypical and functional characterization of different organ systems. Due to the large amount of imaging data, automated image analysis is required, which can be performed using deep learning methods, e. g. for automated organ segmentation. In this paper we describe a computational pipeline for automated segmentation of abdominal organs on MRI data from 20,000 participants of UKBB and NAKO and provide results of the quality control process. We found that approx. 90% of data sets showed no relevant segmentation errors while relevant errors occurred in a varying proportion of data sets depending on the organ of interest. Image-derived features based on automated organ segmentations showed relevant deviations of varying degree in the presence of segmentation errors. These results show that large-scale, deep learning-based abdominal organ segmentation on MRI data is feasible with overall high accuracy, but visual quality control remains an important step ensuring the validity of down-stream analyses in large epidemiological imaging studies. [ABSTRACT FROM AUTHOR]

Clinical coding is the task of transforming medical information in a patient's health records into structured codes so that they can be used for statistical analysis. This is a cognitive and time-consuming task that follows a standard process in order to achieve a high level of consistency. Clinical coding could potentially be supported by an automated system to improve the efficiency and accuracy of the process. We introduce the idea of automated clinical coding and summarise its challenges from the perspective of Artificial Intelligence (AI) and Natural Language Processing (NLP), based on the literature, our project experience over the past two and half years (late 2019–early 2022), and discussions with clinical coding experts in Scotland and the UK. Our research reveals the gaps between the current deep learning-based approach applied to clinical coding and the need for explainability and consistency in real-world practice. Knowledge-based methods that represent and reason the standard, explainable process of a task may need to be incorporated into deep learning-based methods for clinical coding. Automated clinical coding is a promising task for AI, despite the technical and organisational challenges. Coders are needed to be involved in the development process. There is much to achieve to develop and deploy an AI-based automated system to support coding in the next five years and beyond. [ABSTRACT FROM AUTHOR]

Disinformation attacks that make use of social media platforms, e.g., the attacks orchestrated by the Russian "Internet Research Agency" during the 2016 U.S. Presidential election campaign and the 2016 Brexit referendum in the UK, have led to increasing demands from governmental agencies for AI tools that are capable of identifying such attacks in their earliest stages, rather than responding to them in retrospect. This research undertaken on behalf of the Canadian Armed Forces and Department of National Defence. Our ultimate objective is the development of an integrated set of machine-learning algorithms which will mobilize artificial intelligence to identify hostile disinformation activities in "near-real-time." Employing The Dark Crawler, the Posit Toolkit, TensorFlow (Deep Neural Networks), plus the Random Forest classifier and short-text classification programs known as LibShortText and LibLinear, we have analysed a wide sample of social media posts that exemplify the "fake news" that was disseminated by Russia's Internet Research Agency, comparing them to "real news" posts in order to develop an automated means of classification. [ABSTRACT FROM AUTHOR]

Inter-individual variation in fat distribution is increasingly recognized as clinically important but is not routinely assessed in clinical practice, in part because medical imaging has not been practical to deploy at scale for this task. Here, we report a deep learning model trained on an individual's body shape outline—or "silhouette" —that enables accurate estimation of specific fat depots of interest, including visceral (VAT), abdominal subcutaneous (ASAT), and gluteofemoral (GFAT) adipose tissue volumes, and VAT/ASAT ratio. Two-dimensional coronal and sagittal silhouettes are constructed from whole-body magnetic resonance images in 40,032 participants of the UK Biobank and used as inputs for a convolutional neural network to predict each of these quantities. Mean age of the study participants is 65 years and 51% are female. A cross-validated deep learning model trained on silhouettes enables accurate estimation of VAT, ASAT, and GFAT volumes (R2: 0.88, 0.93, and 0.93, respectively), outperforming a comparator model combining anthropometric and bioimpedance measures (ΔR2 = 0.05–0.13). Next, we study VAT/ASAT ratio, a nearly body-mass index (BMI)—and waist circumference-independent marker of metabolically unhealthy fat distribution. While the comparator model poorly predicts VAT/ASAT ratio (R2: 0.17–0.26), a silhouette-based model enables significant improvement (R2: 0.50–0.55). Increased silhouette-predicted VAT/ASAT ratio is associated with increased risk of prevalent and incident type 2 diabetes and coronary artery disease independent of BMI and waist circumference. These results demonstrate that body silhouette images can estimate important measures of fat distribution, laying the scientific foundation for scalable population-based assessment. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy (DR) screening images are heterogeneous and contain undesirable non-retinal, incorrect field and ungradable samples which require curation, a laborious task to perform manually. We developed and validated single and multi-output laterality, retinal presence, retinal field and gradability classification deep learning (DL) models for automated curation. The internal dataset comprised of 7743 images from DR screening (UK) with 1479 external test images (Portugal and Paraguay). Internal vs external multi-output laterality AUROC were right (0.994 vs 0.905), left (0.994 vs 0.911) and unidentifiable (0.996 vs 0.680). Retinal presence AUROC were (1.000 vs 1.000). Retinal field AUROC were macula (0.994 vs 0.955), nasal (0.995 vs 0.962) and other retinal field (0.997 vs 0.944). Gradability AUROC were (0.985 vs 0.918). DL effectively detects laterality, retinal presence, retinal field and gradability of DR screening images with generalisation between centres and populations. DL models could be used for automated image curation within DR screening. [ABSTRACT FROM AUTHOR]

Genome-wide association studies (GWASs) examine the association between genotype and phenotype while adjusting for a set of covariates. Although the covariates may have non-linear or interactive effects, due to the challenge of specifying the model, GWAS often neglect such terms. Here we introduce DeepNull, a method that identifies and adjusts for non-linear and interactive covariate effects using a deep neural network. In analyses of simulated and real data, we demonstrate that DeepNull maintains tight control of the type I error while increasing statistical power by up to 20% in the presence of non-linear and interactive effects. Moreover, in the absence of such effects, DeepNull incurs no loss of power. When applied to 10 phenotypes from the UK Biobank (n = 370K), DeepNull discovered more hits (+6%) and loci (+7%), on average, than conventional association analyses, many of which are biologically plausible or have previously been reported. Finally, DeepNull improves upon linear modeling for phenotypic prediction (+23% on average). GWAS often assume a linear phenotype-covariate relationship which may not hold in practice. Here the authors present DeepNull, in which they apply deep learning to identify and adjust for complex non-linear relationships, improving phenotypic prediction and GWAS power. [ABSTRACT FROM AUTHOR]

Convolutional neural networks (CNNs)—as a type of deep learning—have been specifically designed for highly heterogeneous data, such as natural images. Neuroimaging data, however, is comparably homogeneous due to (1) the uniform structure of the brain and (2) additional efforts to spatially normalize the data to a standard template using linear and non-linear transformations. To harness spatial homogeneity of neuroimaging data, we suggest here a new CNN architecture that combines the idea of hierarchical abstraction in CNNs with a prior on the spatial homogeneity of neuroimaging data. Whereas early layers are trained globally using standard convolutional layers, we introduce patch individual filters (PIF) for higher, more abstract layers. By learning filters in individual latent space patches without sharing weights, PIF layers can learn abstract features faster and specific to regions. We thoroughly evaluated PIF layers for three different tasks and data sets, namely sex classification on UK Biobank data, Alzheimer's disease detection on ADNI data and multiple sclerosis detection on private hospital data, and compared it with two baseline models, a standard CNN and a patch-based CNN. We obtained two main results: First, CNNs using PIF layers converge consistently faster, measured in run time in seconds and number of iterations than both baseline models. Second, both the standard CNN and the PIF model outperformed the patch-based CNN in terms of balanced accuracy and receiver operating characteristic area under the curve (ROC AUC) with a maximal balanced accuracy (ROC AUC) of 94.21% (99.10%) for the sex classification task (PIF model), and 81.24% and 80.48% (88.89% and 87.35%) respectively for the Alzheimer's disease and multiple sclerosis detection tasks (standard CNN model). In conclusion, we demonstrated that CNNs using PIF layers result in faster convergence while obtaining the same predictive performance as a standard CNN. To the best of our knowledge, this is the first study that introduces a prior in form of an inductive bias to harness spatial homogeneity of neuroimaging data. [ABSTRACT FROM AUTHOR]

Peoples' recreation and well-being are closely related to their aesthetic enjoyment of the landscape. Ecosystem service (ES) assessments record the aesthetic contributions of landscapes to peoples' well-being in support of sustainable policy goals. However, the survey methods available to measure these contributions restrict modelling at large scales. As a result, most studies rely on environmental indicator models but these do not incorporate peoples' actual use of the landscape. Now, social media has emerged as a rich new source of information to understand human-nature interactions while advances in deep learning have enabled large-scale analysis of the imagery uploaded to these platforms. In this study, we test the accuracy of Flickr and deep learning-based models of landscape quality using a crowdsourced survey in Great Britain. We find that this novel modelling approach generates a strong and comparable level of accuracy versus an indicator model and, in combination, captures additional aesthetic information. At the same time, social media provides a direct measure of individuals' aesthetic enjoyment, a point of view inaccessible to indicator models, as well as a greater independence of the scale of measurement and insights into how peoples' appreciation of the landscape changes over time. Our results show how social media and deep learning can support significant advances in modelling the aesthetic contributions of ecosystems for ES assessments. [ABSTRACT FROM AUTHOR]