Sleep Apnea Prediction Using Deep Learning.

Obstructive sleep apnea (OSA) is a sleep disorder that causes partial or complete cessation of breathing during an individual's sleep. Various methods have been proposed to automatically detect OSA events, but little work has focused on predicting such events in advance, which is useful for the development of devices that regulate breathing during a patient's sleep. We propose four methods for sleep apnea prediction based on convolutional and long short-term memory neural networks (1D-CNN, ConvLSTM, 1D-CNN-LSTM and 2D-CNN-LSTM), which use raw data from three respiratory signals (nasal flow, abdominal and thoracic) sampled at 32 Hz, without any human-engineered features. We predict OSA (apnea or hypopnea) and normal breathing events 30 seconds ahead using the prior 90 seconds' data. Our results on a dataset containing over 46,000 examples from 1,507 subjects show that all four models achieved promising accuracy ( 81%). The 1D-CNN-LSTM and 2D-CNN-LSTM were the best two performing models with accuracy, sensitivity and specificity over 83%, 81% and 85% respectively. These results show that OSA events can be accurately predicted in advance based on respiratory signals, opening up opportunities for the development of devices to preemptively regulate the airflow to sleepers to avoid these events. Furthermore, we demonstrate good prediction performance even when respiratory signals are downsampled by a factor of 32, to 1 Hz, for which our proposed 1D-CNN-LSTM achieved 82.94% accuracy, 81.25% sensitivity and 84.63% specificity. This robustness to low sampling frequencies allows our algorithms to be implemented in devices with low storage capacity, making them suitable for at-home environments.

TrackSOM: Mapping immune response dynamics through clustering of time-course cytometry data.

Mapping the dynamics of immune cell populations over time or disease-course is key to understanding immunopathogenesis and devising putative interventions. We present TrackSOM, a novel method for delineating cellular populations and tracking their development over a time- or disease-course cytometry datasets. We demonstrate TrackSOM-enabled elucidation of the immune response to West Nile Virus infection in mice, uncovering heterogeneous subpopulations of immune cells and relating their functional evolution to disease severity. TrackSOM is easy to use, encompasses few parameters, is quick to execute, and enables an integrative and dynamic overview of the immune system kinetics that underlie disease progression and/or resolution.

The Contribution of Nutrients of Concern to the Diets of 18-to-30-Year-Old Australians from Food Prepared Outside Home Differs by Food Outlet Types: The MYMeals Cross-Sectional Study.

Young adults are frequent consumers of food prepared outside the home (FOH). In a cross-sectional survey, the MYMeals study, we showed FOH provided one-third of meals and snacks for young Australian adults, yet it contributed higher proportions of energy and nutrients of concern, such as saturated fat and sodium. This study aimed to determine the detailed proportional contribution of nutrients of concern from the nine food outlet types captured in the MYMeals study. Young adults residing in New South Wales (NSW), Australia, (n = 1001) used a validated smartphone app to report all types and amounts of food and beverages consumed for three consecutive days, as well as their preparation location. The proportions of daily energy, macronutrients, sodium, total sugars, and saturated fat were calculated for each of the nine following outlet types: bakeries or patisseries, coffee chains, cold-drink chains, fast-food chains, ice creamery or frozen yoghurt outlets, independent cafes or restaurants, pubs (hotels) and clubs, service stations or convenience stores, and others not fitting the above categories. Of all FOH outlet types, independent cafes or restaurants contributed the most energy (17.5%), sodium (20.0%) and saturated fat (17.8%) to the total diet, followed by fast-food chains (12.0% energy, 15.8% sodium, and 12.0% saturated fat) and other outlets, with smaller proportions. For males, the proportion of energy and nutrients contributed by fast-food outlets was higher than for females (14.8% versus 9.8% energy). Menu labelling at independent cafes and restaurants is recommended, comprising, in addition to the energy labels already in use in fast-food restaurants, the labelling of nutrients of concern. The feasibility of this recommendation warrants further exploration.

Association of Longitudinal Sleep and Next-day Indoor Mobility Measured via Passive Sensors among Community-dwelling Older Adults.

Previous studies have shown there is a relationship between sleep and mobility in older adults by collecting and analysing self-reported data from surveys and questionnaires, or by using objective measures from polysomnography or actigraphy. However, these methods have limitations for long-term monitoring, especially for community-dwelling adults. In this paper, we investigate the association between sleep and indoor mobility using longitudinal data collected over a period of about 12 months for older adults (65 years or older) living at home in Australia. The data was collected objectively and continuously using non-invasive and passive sensors. First, we explored whether sleep and indoor mobility are different across gender and age groups (70s, 80s, and 90s). Second, we investigate the association of sleep and next-day indoor mobility through a stepwise multivariate regression. We found that males and females have significant differences in mobility, time in bed, total time in sleep, number and duration of awakenings and sleep efficiency. Additionally, mobility and all sleep measures significantly vary across the three age groups, except for sleep onset latency between 80s and 90s. Our findings show that sleep efficiency and total sleep time are the key sleep measures affecting next-day mobility, while sleep onset latency has the least effect.Clinical relevance - Our study contributes to a better understanding of the sleep patterns of older adults and how they affect their physical functioning.

The Contribution of Foods Prepared Outside the Home to the Diets of 18- to 30-Year-Old Australians: The MYMeals Study.

Young adults are the highest consumers of food prepared outside home (FOH) and gain most weight among Australian adults. One strategy to address the obesogenic food environment is menu labelling legislation whereby outlets with >20 stores in one state and >50 Australia-wide must display energy content in kJ. The aim of this study was to assess the contribution of FOH to the energy and macronutrients, saturated fat, total sugars and sodium intakes of young Australians. One thousand and one 18 to 30-year-olds (57% female) residing in Australia's most populous state recorded all foods and beverages consumed and the location of preparation for three consecutive days using a purpose-designed smartphone application. Group means for the daily consumption of energy, percentage energy (%E) for protein, carbohydrate, total sugars, total and saturated fats, and sodium density (mg/1000 kJ) and proportions of nutrients from FOH from menu labelling and independent outlets were compared. Overall, participants consumed 42.4% of their energy intake from FOH with other nutrients ranging from 39.8% (sugars) to 47.3% (sodium). Independent outlets not required to label menus, contributed a greater percentage of energy (23.6%) than menu labelling outlets (18.7%, p < 0.001). Public health policy responses such as public education campaigns, extended menu labelling, more detailed nutrition information and reformulation targets are suggested to facilitate healthier choices.

Using single-cell cytometry to illustrate integrated multi-perspective evaluation of clustering algorithms using Pareto fronts.

MOTIVATION: Many 'automated gating' algorithms now exist to cluster cytometry and single cell sequencing data into discrete populations. Comparative algorithm evaluations on benchmark datasets rely either on a single performance metric, or a few metrics considered independently of one another. However, single metrics emphasise different aspects of clustering performance and do not rank clustering solutions in the same order. This underlies the lack of consensus between comparative studies regarding optimal clustering algorithms and undermines the translatability of results onto other non-benchmark datasets. RESULTS: We propose the Pareto fronts framework as an integrative evaluation protocol, wherein individual metrics are instead leveraged as complementary perspectives. Judged superior are algorithms that provide the best trade-off between the multiple metrics considered simultaneously. This yields a more comprehensive and complete view of clustering performance. Moreover, by broadly and systematically sampling algorithm parameter values using the Latin Hypercube sampling method, our evaluation protocol minimises (un)fortunate parameter value selections as confounding factors. Furthermore, it reveals how meticulously each algorithm must be tuned in order to obtain good results, vital knowledge for users with novel data. We exemplify the protocol by conducting a comparative study between three clustering algorithms (ChronoClust, FlowSOM and Phenograph) using four common performance metrics applied across four cytometry benchmark datasets. To our knowledge, this is the first time Pareto fronts have been used to evaluate the performance of clustering algorithms in any application domain. AVAILABILITY: Implementation of our Pareto front methodology and all scripts to reproduce this article are available at https://github.com/ghar1821/ParetoBench.

Detecting Personality Traits Using Inter-Hemispheric Asynchrony of the Brainwaves.

Affective personality traits have been associated with a risk of developing mental and cognitive disorders and can be informative for early detection and management of such disorders. However, conventional personality trait detection is often biased and unreliable, as it depends on the honesty of the subjects when filling out the lengthy questionnaires. In this paper, we propose a method for objective detection of personality traits using physiological signals. Subjects are shown affective images and videos to evoke a range of emotions. The electrical activity of the brain is captured using EEG during this process and the multi-channel EEG data is processed to compute the inter-hemispheric asynchrony of the brainwaves. The most discriminative features are selected and then used to build a machine learning classifier, which is trained to predict 16 personality traits. Our results show high predictive accuracy for both image and video stimuli individually, and an improvement when the two stimuli are combined, achieving a 95.49% accuracy. Most of the selected discriminative features were found to be extracted from the alpha frequency band. Our work shows that personality traits can be accurately detected with EEG data, suggesting possible use in practical applications for early detection of mental and cognitive disorders.

The Sydney Triage to Admission Risk Tool (START2) using machine learning techniques to support disposition decision-making.

OBJECTIVE: To further develop and refine an Emergency Department (ED) in-patient admission prediction model using machine learning techniques. METHODS: This was a retrospective analysis of state-wide ED data from New South Wales, Australia. Six classification algorithms (Bayesian networks, decision trees, logistic regression, naïve Bayes, neural networks and nearest neighbour) and five feature selection techniques (none, manual, correlation-based, information gain and wrapper) were examined. Presenting problem was categorised using broad (n = 20) and specific (n = 100) representations. Models were evaluated based on Area Under the Curve (AUC) and accuracy. The results were compared with the Sydney Triage to Admission Risk Tool (START), which uses logistic regression and six manually selected features. RESULTS: Sixty admission prediction models were trained and validated using data from 1 721 294 patients. Under the broad representation of presenting problem, the nearest neighbour algorithm with manual feature selection had the best AUC of 0.8206 (95% CI ±0.0006), while the decision tree with no feature selection had the best accuracy of 74.83% (95% CI ±0.065). Under the specific representation, almost all models improved; the nearest neighbour with information gain feature selection had the best AUC of 0.8267 (95% CI ±0.0006), while the decision tree with wrapper or no feature selection had the best accuracy of 75.24% (95% CI ±0.064). Eleven of the machine learning models had slightly better AUC than the START model. CONCLUSION: Machine learning methods demonstrate similar performance to logistic regression for ED disposition prediction models using basic triage information. This should be investigated further, especially for larger data sets with more complex clinical information.

Examining the Frequency and Contribution of Foods Eaten Away From Home in the Diets of 18- to 30-Year-Old Australians Using Smartphone Dietary Assessment (MYMeals): Protocol for a Cross-Sectional Study.

BACKGROUND: Young Australians aged between 18 and 30 years have experienced the largest increase in the body mass index and spend the largest proportion of their food budget on fast food and eating out. Frequent consumption of foods purchased and eaten away from home has been linked to poorer diet quality and weight gain. There has been no Australian research regarding quantities, type, or the frequency of consumption of food prepared outside the home by young adults and its impact on their energy and nutrient intakes. OBJECTIVES: The objective of this study was to determine the relative contributions of different food outlets (eg, fast food chain, independent takeaway food store, coffee shop, etc) to the overall food and beverage intake of young adults; to assess the extent to which food and beverages consumed away from home contribute to young adults' total energy and deleterious nutrient intakes; and to study social and physical environmental interactions with consumption patterns of young adults. METHODS: A cross-sectional study of 1008 young adults will be conducted. Individuals are eligible to participate if they: (1) are aged between 18 and 30 years; (2) reside in New South Wales, Australia; (3) own or have access to a smartphone; (4) are English-literate; and (5) consume at least one meal, snack, or drink purchased outside the home per week. An even spread of gender, age groups (18 to 24 years and 25 to 30 years), metropolitan or regional geographical areas, and high and low socioeconomic status areas will be included. Participants will record all food and drink consumed over 3 consecutive days, together with location purchased and consumed in our customized smartphone app named Eat and Track (EaT). Participants will then complete an extensive demographics questionnaire. Mean intakes of energy, nutrients, and food groups will be calculated along with the relative contribution of foods purchased and eaten away from home. A subsample of 19.84% (200/1008) of the participants will complete three 24-hour recall interviews to compare with the data collected using EaT. Data mining techniques such as clustering, decision trees, neural networks, and support vector machines will be used to build predictive models and identify important patterns. RESULTS: Recruitment is underway, and results will be available in 2018. CONCLUSIONS: The contribution of foods prepared away from home, in terms of energy, nutrients, deleterious nutrients, and food groups to young people's diets will be determined, as will the impact on meeting national recommendations. Foods and consumption behaviors that should be targeted in future health promotion efforts for young adults will be identified.

Automated Classification of Clinical Incident Types.

We consider the task of automatic classification of clinical incident reports using machine learning methods. Our data consists of 5448 clinical incident reports collected from the Incident Information Management System used by 7 hospitals in the state of New South Wales in Australia. We evaluate the performance of four classification algorithms: decision tree, naïve Bayes, multinomial naïve Bayes and support vector machine. We initially consider 13 classes (incident types) that were then reduced to 12, and show that it is possible to build accurate classifiers. The most accurate classifier was the multinomial naïve Bayes achieving accuracy of 80.44% and AUC of 0.91. We also investigate the effect of class labelling by an ordinary clinician and an expert, and show that when the data is labelled by an expert the classification performance of all classifiers improves. We found that again the best classifier was multinomial naïve Bayes achieving accuracy of 81.32% and AUC of 0.97. Our results show that some classes in the Incident Information Management System such as Primary Care are not distinct and their removal can improve performance; some other classes such as Aggression Victim are easier to classify than others such as Behavior and Human Performance. In summary, we show that the classification performance can be improved by expert class labelling of the training data, removing classes that are not well defined and selecting appropriate machine learning classifiers.