How to treat mixed behavior segments in supervised machine learning of behavioural modes from inertial measurement data.

The application of supervised machine learning methods to identify behavioural modes from inertial measurements of bio-loggers has become a standard tool in behavioural ecology. Several design choices can affect the accuracy of identifying the behavioural modes. One such choice is the inclusion or exclusion of segments consisting of more than a single behaviour (mixed segments) in the machine learning model training data. Currently, the common practice is to ignore such segments during model training. In this paper we tested the hypothesis that including mixed segments in model training will improve accuracy, as the model would perform better in identifying them in the test data. We test this hypothesis using a series of data simulations on four datasets of accelerometer data coupled with behaviour observations, obtained from four study species (Damaraland mole-rats, meerkats, olive baboons, polar bears). Results show that when a substantial proportion of the test data are mixed behaviour segments (above ~ 10%), including mixed segments in machine learning model training improves the accuracy of classification. These results were consistent across the four study species, and robust to changes in segment length, sample size, and degree of mixture within the mixed segments. However, we also find that in some cases (particularly in baboons) models trained with mixed segments show reduced accuracy in classifying test data containing only single behaviour (pure) segments, compared to models trained without mixed segments. Based on these results, we recommend that when the classification model is expected to deal with a substantial proportion of mixed behaviour segments (> 10%), it is beneficial to include them in model training, otherwise, it is unnecessary but also not harmful. The exception is when there is a basis to assume that the training data contains a higher rate of mixed segments than the actual (unobserved) data to be classified-such a situation may occur particularly when training data are collected in captivity and used to classify data from the wild. In this case, excess inclusion of mixed segments in training data should probably be avoided.

Early-life behaviour predicts first-year survival in a long-distance avian migrant.

Early-life conditions have critical, long-lasting effects on the fate of individuals, yet early-life activity has rarely been linked to subsequent survival of animals in the wild. Using high-resolution GPS and body-acceleration data of 93 juvenile white storks (Ciconia ciconia), we examined the links between behaviour during both pre-fledging and post-fledging (fledging-to-migration) periods and subsequent first-year survival. Juvenile daily activity (based on overall dynamic body acceleration) showed repeatable between-individual variation, the juveniles' pre- and post-fledging activity levels were correlated and both were positively associated with subsequent survival. Daily activity increased gradually throughout the post-fledging period, and the relationship between post-fledging activity and survival was stronger in individuals who increased their daily activity level faster (an interaction effect). We suggest that high activity profiles signified individuals with increased pre-migratory experience, higher individual quality and perhaps more proactive personality, which could underlie their superior survival rates. The duration of individuals' fledging-to-migration periods had a hump-shaped relationship with survival: higher survival was associated with intermediate rather than short or long durations. Short durations reflect lower pre-migratory experience, whereas very long ones were associated with slower increases in daily activity level which possibly reflects slow behavioural development. In accordance with previous studies, heavier nestlings and those that hatched and migrated earlier had increased survival. Using extensive tracking data, our study exposed new links between early-life attributes and survival, suggesting that early activity profiles in migrating birds can explain variation in first-year survival.

Promoting head CT exams in the emergency department triage using a machine learning model.

PURPOSE: In this study, we aimed to develop a novel prediction model to identify patients in need of a non-contrast head CT exam during emergency department (ED) triage. METHODS: We collected data of all adult ED visits in our institution for five consecutive years (1/2013-12/2017). Retrieved variables included the following: demographics, mode of arrival to the ED, comorbidities, home medications, structured and unstructured chief complaints, vital signs, pain scale score, emergency severity index, ED wing assignment, documentation of previous ED visits, hospitalizations and CTs, and current visit non-contrast head CT usage. A machine learning gradient boosting model was trained on data from the years 2013-2016 and tested on data from 2017. Area under the curve (AUC) was used as metrics. Single-variable AUCs were also determined. Youden's index evaluated optimal sensitivity and specificity of the models. RESULTS: The final cohort included 595,561 ED visits. Non-contrast head CT usage rate was 11.8%. Each visit was coded into an input vector of 171 variables. Single-variable analysis showed that chief complaint had the best single predictive analysis (AUC = 0.87). The best model showed an AUC of 0.93 (95% CI 0.931-0.936) for predicting non-contrast head CT usage at triage level. The model had a sensitivity of 88.1% and specificity of 85.7% for non-contrast head CT utilization. CONCLUSION: The developed model can identify patients that need to undergo head CT exam already in the ED triage level and by that allow faster diagnosis and treatment.

A Gradient Boosting Machine Learning Model for Predicting Early Mortality in the Emergency Department Triage: Devising a Nine-Point Triage Score.

BACKGROUND: Emergency departments (ED) are becoming increasingly overwhelmed, increasing poor outcomes. Triage scores aim to optimize the waiting time and prioritize the resource usage. Artificial intelligence (AI) algorithms offer advantages for creating predictive clinical applications. OBJECTIVE: Evaluate a state-of-the-art machine learning model for predicting mortality at the triage level and, by validating this automatic tool, improve the categorization of patients in the ED. DESIGN: An institutional review board (IRB) approval was granted for this retrospective study. Information of consecutive adult patients (ages 18-100) admitted at the emergency department (ED) of one hospital were retrieved (January 1, 2012-December 31, 2018). Features included the following: demographics, admission date, arrival mode, referral code, chief complaint, previous ED visits, previous hospitalizations, comorbidities, home medications, vital signs, and Emergency Severity Index (ESI). The following outcomes were evaluated: early mortality (up to 2 days post ED registration) and short-term mortality (2-30 days post ED registration). A gradient boosting model was trained on data from years 2012-2017 and examined on data from the final year (2018). The area under the curve (AUC) for mortality prediction was used as an outcome metric. Single-variable analysis was conducted to develop a nine-point triage score for early mortality. KEY RESULTS: Overall, 799,522 ED visits were available for analysis. The early and short-term mortality rates were 0.6% and 2.5%, respectively. Models trained on the full set of features yielded an AUC of 0.962 for early mortality and 0.923 for short-term mortality. A model that utilized the nine features with the highest single-variable AUC scores (age, arrival mode, chief complaint, five primary vital signs, and ESI) yielded an AUC of 0.962 for early mortality. CONCLUSION: The gradient boosting model shows high predictive ability for screening patients at risk of early mortality utilizing data available at the time of triage in the ED.

Unilesional mycosis fungoides is associated with increased expression of microRNA-17~92 and T helper 1 skewing.

BACKGROUND: The molecular basis of unilesional mycosis fungoides (MF), characterized by a solitary lesion that is clinicopathologically indistinguishable from multifocal patch or plaque MF (early MF), is unknown. OBJECTIVES: To investigate the microRNA profile in unilesional MF distinguishing it from early MF. METHODS: Biopsy samples of unilesional MF and early MF were evaluated with the Affymetrix microRNA array, with further comparison with inflammatory dermatosis, using quantitative polymerase chain reaction. NanoString technology was applied to analyse the gene expression of T helper (Th)1 immune markers, and immunohistochemistry was used to evaluate CXCR3 and GATA-binding protein 3 (GATA3) markers for Th1 and Th2 cells, respectively. RESULTS: Unilesional MF had a significantly higher level of expression of all members of the microRNA miR-17~92 cluster than early MF. Specifically, unilesional MF had a higher miR-17 level than early MF and inflammatory dermatoses. There was downregulation of the expression of phosphatase and tensin homolog (PTEN) and CREB1, known targets of miR-17~92 members; higher gene expression of interleukin-2 and interferon-γ; and a statistically lower average percentage of GATA3+ dermal cells (6·7% vs. 42·3%), were detected in unilesional MF compared with early MF. High immunoreactivity of CXCR3 was noted in both unilesional and early MF. CONCLUSIONS: Unilesional MF exhibits a microRNA profile distinct from that of conventional early MF, with a higher level of miR-17~92 members along with Th1 skewing. These findings suggest a robust reactive T-cell immune response in unilesional MF and might account for the localized nature of this disease.

Oncogenic role of microRNA-155 in mycosis fungoides: an in vitro and xenograft mouse model study.

BACKGROUND: MicroRNA (miR)-155 contributes to the proliferation of mycosis fungoides (MF) in vitro and is upregulated in tumours of MF compared with early MF lesions. OBJECTIVES: To investigate the contribution of miR-155 to the cancerous phenotype and drug resistance of MF/Sézary cell lines. METHODS: miR-155 was inhibited in MF cell lines (MyLa and MJ) by transduction of miRZip anti-miR-155, and overexpressed in Hut78 cells by transduction of miRVec-miR-155; empty plasmids served as controls. Cells were analysed for response to inducers of apoptosis and cell-cycle arrest, using fluorescence-activated cell sorting. Transduced MyLa cells were subcutaneously injected into severe combined immunodeficient mice, and tumours were analysed immunohistochemically and for final size. RESULT: MyLa and MJ cells expressed a high level of miR-155; Hut78 cells expressed a low level. MF cell lines stably expressing miR-155 inhibitor showed increased G2/M arrest in response to N-p-tolyl-2-(3,4,5-trimethoxyphenyl quinazolin-4-amine) (SL111), an inducer of cell-cycle arrest, followed by increased apoptosis. Additionally, they showed increased apoptosis in response to suberoylanilide hydroxamic acid (SAHA). Tumours formed in mice from injected anti-miR-155-expressing MyLa cells had a significantly lower volume and higher occurrence of apoptosis than controls. Stable overexpression of miR-155 in Hut78 cells had no effect. CONCLUSIONS: Oncogenic miR-155 appears to contribute to the cancerous phenotype of MyLa and MJ cells, but not of Hut78 cells, by interrupting activation of the G2/M checkpoint in response to SL111, and decreasing apoptosis in response to SL111 and SAHA, thereby facilitating tumour growth. These findings have implications for the potential development of novel therapeutic modalities for MF incorporating miR-155 inhibitors.

Using accelerometry to compare costs of extended migration in an arctic herbivore.

Understanding how individuals manage costs during the migration period is challenging because individuals are difficult to follow between sites; the advent of hybrid Global Positioning System-acceleration (ACC) tracking devices enables researchers to link spatial and temporal attributes of avian migration with behavior for the first time ever. We fitted these devices on male Greenland white-fronted geese Anser albifrons flavirostris wintering at 2 sites (Loch Ken, Scotland and Wexford, Ireland) to understand whether birds migrating further during spring fed more on wintering and staging areas in advance of migration episodes. Although Irish birds flew significantly further (ca. 300 km) than Scottish birds during spring, their cumulative hours of migratory flight, flight speed during migration, and overall dynamic body ACC (i.e., a proxy for energy expenditure) were not significantly different. Further, Irish birds did not feed significantly more or expend significantly more energy in advance of migration episodes. These results suggest broad individual plasticity in this species, although Scottish birds arriving on breeding areas in Greenland with greater energy stores (because they migrated less) may be better prepared for food scarcity, which might increase their reproductive success.

The challenges of the first migration: movement and behaviour of juvenile vs. adult white storks with insights regarding juvenile mortality.

Migration conveys an immense challenge, especially for juvenile birds coping with enduring and risky journeys shortly after fledging. Accordingly, juveniles exhibit considerably lower survival rates compared to adults, particularly during migration. Juvenile white storks (Ciconia ciconia), which are known to rely on adults during their first fall migration presumably for navigational purposes, also display much lower annual survival than adults. Using detailed GPS and body acceleration data, we examined the patterns and potential causes of age-related differences in fall migration properties of white storks by comparing first-year juveniles and adults. We compared juvenile and adult parameters of movement, behaviour and energy expenditure (estimated from overall dynamic body acceleration) and placed this in the context of the juveniles' lower survival rate. Juveniles used flapping flight vs. soaring flight 23% more than adults and were estimated to expend 14% more energy during flight. Juveniles did not compensate for their higher flight costs by increased refuelling or resting during migration. When juveniles and adults migrated together in the same flock, the juvenile flew mostly behind the adult and was left behind when they separated. Juveniles showed greater improvement in flight efficiency throughout migration compared to adults which appears crucial because juveniles exhibiting higher flight costs suffered increased mortality. Our findings demonstrate the conflict between the juveniles' inferior flight skills and their urge to keep up with mixed adult-juvenile flocks. We suggest that increased flight costs are an important proximate cause of juvenile mortality in white storks and likely in other soaring migrants and that natural selection is operating on juvenile variation in flight efficiency.

AcceleRater: a web application for supervised learning of behavioral modes from acceleration measurements.

BACKGROUND: The study of animal movement is experiencing rapid progress in recent years, forcefully driven by technological advancement. Biologgers with Acceleration (ACC) recordings are becoming increasingly popular in the fields of animal behavior and movement ecology, for estimating energy expenditure and identifying behavior, with prospects for other potential uses as well. Supervised learning of behavioral modes from acceleration data has shown promising results in many species, and for a diverse range of behaviors. However, broad implementation of this technique in movement ecology research has been limited due to technical difficulties and complicated analysis, deterring many practitioners from applying this approach. This highlights the need to develop a broadly applicable tool for classifying behavior from acceleration data. DESCRIPTION: Here we present a free-access python-based web application called AcceleRater, for rapidly training, visualizing and using models for supervised learning of behavioral modes from ACC measurements. We introduce AcceleRater, and illustrate its successful application for classifying vulture behavioral modes from acceleration data obtained from free-ranging vultures. The seven models offered in the AcceleRater application achieved overall accuracy of between 77.68% (Decision Tree) and 84.84% (Artificial Neural Network), with a mean overall accuracy of 81.51% and standard deviation of 3.95%. Notably, variation in performance was larger between behavioral modes than between models. CONCLUSIONS: AcceleRater provides the means to identify animal behavior, offering a user-friendly tool for ACC-based behavioral annotation, which will be dynamically upgraded and maintained.