Preliminary Dataset Research Articles

Intra-examination agreement between multi-echo gradient echo and confounder-corrected chemical shift-encoded MR sequences for R2* estimation as a biomarker of liver iron content in patients with a wide range of T2*/R2* and proton density fat fraction values.

To investigate the intra-examination agreement between multi-echo gradient echo (MEGE) and confounder-corrected chemical shift-encoded (CSE) sequences for liver T2*/R2* estimations in a wide range of T2*/R2* and proton density fat fraction (PDFF) values. Exploratorily, to search for the T2*/R2* value where the agreement line breaks and examine differences between regions of low and high agreement. Consecutive patients at risk for liver iron overload who underwent MEGE and CSE sequences within the same exam at 1.5T were retrospectively selected. Regions of interest were drawn in the right and one in the left liver lobes on post-processed images for R2*(sec-1) and PDFF (%) estimation. Agreement between MEGE-R2* and CSE-R2* was evaluated using intra-class correlation coefficient (ICC) and Bland-Altman analysis. 95% confidence intervals (CI) were computed. Segment-and-regression analysis was performed to find the point where the agreement between sequences is interrupted. Regions of low and high agreement were examined using tree-based partitioning analyses. 49 patients were included. Mean MEGE-R2* was 94.2s-1 (range: 31.0-737.1) and mean CSE-R2* 87.7 (29.7-748.1). Mean CSE-PDFF was 9.12% (0.1-43.3). Agreement was strong for R2* estimations (ICC: 0.992,95%CI 0.987,0.996), but the relation was nonlinear and possibly heteroskedastic. Lower agreement occurred when MEGE-R2* > 235s-1, with MEGE-R2* values consistently lower than CSE-R2*. Higher agreement was observed when PDFF < 14%. MEGE-R2* and CSE-R2* strongly agree, though at higher iron content, MEGE-R2* is consistently lower than CSE-R2*. In this preliminary dataset, a breaking point for agreement was found at R2* > 235. Lower agreement was observed in patients with moderate to severe liver steatosis.

5613275 USE OF CONSUMER WEARABLES TO PREDICT PAIN IN PATIENTS WITH SICKLE CELL DISEASE

Background: Sickle cell disease (SCD) is associated with increased morbidity and mortality. Episodes of acute and severe pain known as vaso-occlusive crises (VOC) are the most common cause for hospital admission. Mobile health (mHealth) has developed promising, patient-friendly, minimally invasive tools to monitor patients remotely. Our previous work has leveraged data from mHealth apps and wearable devices to evaluate several machine learning (ML) models to accurately predict pain scores in patients with SCD currently admitted for VOC to the SCD Day Hospital. Aims: Evaluate the feasibility of extended monitoring for 30 days post-discharge and to refine the development of ML models to predict pain scores in patients with SCD. Methods: Patients with SCD aged 18 and above who were admitted for a VOC to the SCD Day Hospital or to Duke University Hospital were eligible for this study. All patients were followed for 30 days following discharge. Following informed consent, patients were provided: 1) a mobile app (Nanpar); and 2) an Apple Watch. Patients were instructed to report their pain to nurses as per standard of care monitoring and at least once daily within the Nanpar app. Patients were asked to continuously wear the Apple Watch, removing only to charge. Physiological data collected by the Apple Watch included heart rate, heart rate variability, oxygen saturation and step count. These data were associated with self-reported pain scores to fit five different machine learning classification models. The performance of the models is evaluated by the following metrics: accuracy, F1-score, root mean squared error (RMSE) and area under the ROC curve (AUC). Results: Nineteen patients were included in this study from April through June 2022. The median age at time of inclusion was 30 years (IQR:22-34). The majority of the patients had genotype HbSS (68%) and all were Black or African American. Eleven patients (55%) were enrolled from the SCD Day Hospital. This preliminary dataset consisted of 1480 data points. After micro-averaging due to the imbalanced dataset, the performance of all the models were very similar. The metrics of the best performing model, the random forest model, were: micro-averaged accuracy:0.89, micro-averaged F1-score:0.50, RMSE:1.52, AUC:0.83. There was no correlation between any of the data elements recorded by the Apple Watch. Our random forest model was able to accurately predict higher pain scores not only for patients who were admitted to the hospital, but also for patients after discharge from the hospital. Discussion: Our model was able to predict pain using data from the Apple Watch with high accuracy, but with low F1-score. The consistency in the performance of each model, along with the low F1-scores reflects the high degree of class imbalance and lack of data most probably due to inconsistent pain reporting after discharge. To alleviate this issue of class imbalance, different oversampling approaches and generative models need to be examined. Future efforts will focus on larger numbers of patients and monitoring patients for longer periods of time to provide a larger dataset, in an attempt to further improve the accuracy of pain prediction. Conclusion: The consumer wearable Apple Watch was a feasible method to collect physiologic data and provided accuracy in prediction of higher pain scores. We believe mHealth efforts can provide valuable insights for patient monitoring and pain prediction for patients with SCD.

Abstract P4-03-43: Disease characteristics and outcomes of people with metastatic breast cancer in a single center cohort study: The Dallas Metastatic Breast Cancer Study

Abstract Background: Breast cancer is the most common cancer in women, and metastatic disease accounts for most breast cancer related deaths. Identifying risk factors for the onset and progression of metastatic breast cancer (MBC) can help us understand how to address and improve morbidity and mortality due to MBC. Large national databases, such the Surveillance, Epidemiology, and End Results (SEER) Program are limited in their ability to capture granular details from patients’ cancer histories. The Dallas Metastatic Breast Cancer Study (DMBCS) is a clinical database established at a single academic medical system to track patient demographics, associated pathology, treatments, and other variables to improve outcomes for patients with MBC. Methods: The DMBCS database was generated from a registry of breast cancer patients submitted to the National Cancer Institute from 2010 to 2021. Patients were initially excluded if they were identified as non-metastatic from this list. Chart review was then done to identify MBC patients from this subsequent group. Demographics, clinical history, pathologic features, lines of treatment, and subsequent recurrences were collected for all patients. Clinical data was stored in REDCap, a secure data collection platform for entering and managing data. Results: 230 cases were included in this preliminary data set. Of the 230 cases, 185 cases were metastatic at time of diagnosis while 45 cases were metastatic recurrences of breast cancer. At diagnosis, 14.8% were less than 40 years of age, 17.3% were 40-49, 35.7% were 50-59, 18.7% were 60-69, and 13.5% were greater than 70 years of age. In terms of ethnicity, 63.5% were White, 27.4% were Black, and 21.7% identified as Hispanic. At time of diagnosis, 30.9% of patients had a BMI classified as overweight and 40.4% were considered obese. Medical comorbidities included hypertension in 43% of cases, diabetes (18.3%), hyperlipidemia (27.4%), and autoimmune disease (13.0%). Clinical subtype analysis revealed 57.8% of patients were hormone receptor positive, 24.8% HER2 positive, and 17.4% triple negative, at diagnosis. 14.8% of cases were diagnosed as inflammatory breast cancer. The most common site of metastasis at presentation was bone with 69.1% of cases, followed by lung (33.9%), liver (30.0%), and brain (15.2%). Over 90% of patients were treated with at least one antineoplastic regimen and 39.6% underwent at least 4 therapies. Calculated 1-year survival after diagnosis was 85.7%. Conclusions: The introduction of the DMBCS will allow continued investigation into clinical drivers of MBC. In this first cohort of patients, we characterized key, yet often underreported, variables and outcomes. Potential applications of this database include investigating the association between obesity and overall survival in patients with MBC, understanding how socioeconomic disparities affects outcomes of patients with MBC, and exploring correlations between autoimmune disease and the progression of MBC. Citation Format: Meng Cao, Mir Lim, Anna Moscowitz, Jonathan Ladner, Christine Hodgdon, Julia Maues, Sangeetha Reddy, Isaac Chan. Disease characteristics and outcomes of people with metastatic breast cancer in a single center cohort study: The Dallas Metastatic Breast Cancer Study. [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P4-03-43.

Towards fast surrogate models for interpolation of tokamak edge plasmas

One of the major design limitations for tokamak fusion reactors is the heat load that can be sustained by the materials at the divertor target. Developing a full understanding of how machine or operation parameters affect the conditions at the divertor requires an enormous number of simulations. A promising approach to circumvent this is to use machine learning models trained on simulation data as surrogate models. Once trained such surrogate models can make fast predictions for any scenario in the design parameter space. In future such simulation based surrogate models could be used in system codes for rapid design studies of future fusion power plants. This work presents the first steps towards the development of such surrogate models for plasma exhaust and the datasets required for their training. Machine learning models like neural networks usually require several thousand data points for training, but the exact amount of data required varies from case to case. Due to the long runtimes of simulations we aim at finding the minimal amount of training data required. A preliminary dataset based on SOLPS-ITER simulations with varying tokamak design parameters, including the major radius, magnetic field strength and neutral density is constructed. To be able to generate more training data within reasonable computation time the simulations in the dataset use fluid neutral simulations and no fluid drift effects. The dataset is used to train a simple neural network and Gradient Boosted Regression Trees and test how the performance depends on the number of training simulations.

Authentication and Geographical Characterisation of Italian Grape Musts through Glucose and Fructose Carbon Isotopic Ratios Determined by LC-IRMS.

The authenticity of grape musts is normally checked through a time-consuming stable isotopic analysis of carbon (δ13C) after fermentation and distillation by following the official OIV MA AS-312-06 method. In this study, the alternative use of a technique based on δ13C isotopic analysis of the major sugars of the grape must by liquid chromatography coupled with isotope ratio mass spectrometry (LC-IRMS) is provided. It allows not only the detection of the fraudulent addition to grape must of exogenous glucose and fructose deriving from C4 plants but also the characterisation of it based on its geographical origin. In order to discriminate between musts from different areas of Italy, a preliminary dataset was considered; the δ13C isotopic ratios of glucose and fructose of around 100 authentic samples were analysed. The two analysed parameters, ranging from -29.8‱ to -21.9‱, are well correlated (R2 = 0.7802) and the northern regions showed significantly more negative δ13C values for both sugars than the rest of the dataset.

An Heuristic Prediction Method for Managing Environmental Blast Noise Impacts

The aim of this work is to manage adverse environmental impacts from long-range blast noise. The work was carried out as part of ongoing research at the DNV Spadeadam Testing and Research site (STaR). STaR carries out crucial major hazards work including improving safety concerns within industry decarbonization sectors and government agencies. The site performs a variety of explosives testing, resulting in environmental blast noise at off-site residential locations. The site is surrounded by complex topography, with terrain featuring range-dependent ground impedance and thermal properties which in turn effects the local meteorology. While accurately modelling blast wave propagation through such environments using traditional computational methods is a computationally expensive task, the required complex and rapidly varying meteorological data are not adequately available. To address this deficiency, a data-driven heuristic method is proposed for the prediction of blast noise levels at several sensitive receivers ranging from 4-14km. The model is formed from a preliminary dataset of off-site blast noise measurements, correlated with a multivariate array of available meteorological data. A principal component analysis is used to determine the atmospheric features which are most influential to sound propagation, and predict the likely range of peak sound pressure levels expected. It is concluded that useful predictions over time scales from an hour to a number of days can be obtained for managing environmental blast noise impacts, and that further measurements of blast noise, along with further correlations with measured atmospheric conditions, could improve the performance of the model.

Determining the Best First-line Modality for Identifying and Localizing Skull Base CSF Leaks

Background: Patients with skull base cerebrospinal fluid (CSF) leaks present with headaches, fluid in the ear(s), and/or rhinorrhea, which are vague symptoms. Beta-2-transferrin protein assays are the gold standard for identifying CSF leaks, but adequate samples cannot always be collected, and the results give no specific localizing information. Medical imaging, including Computed Tomography (CT), Magnetic Resonance (MR), and Nuclear Medicine (NM) cisternography can be utilized to identify and localize CSF leaks, but are imperfect tests. This study aims to determine the best imaging modality for identifying and localizing skull base CSF leaks by comparing CT, MR, and NM Cisternogram results to beta-2-transferrin assay and intraoperative visualization as criterion standards.&#x0D; Methods: In this ongoing study, patient cisternogram and pre-cisternogram imaging and results were acquired from the electronic medical record, radiology information system, and picture archiving and communication system. Inclusion criteria include age greater than 18, suspected skull base CSF leak, and CT, MR, and/or NM cisternogram performed. MR cisternogram procedure included intrathecal gadolinium and NM SPECT-CT cisternogram procedure included intrathecal Tc-99m Sulfur Colloid. These data will be analyzed for positive and negative predictive values, sensitivity, specificity, and ROC Curve comparing to beta-2-transferrin assay results and surgical findings.&#x0D; Results: From 2018-2022, 30 patients with suspected skull base CSF leaks were evaluated. In this limited preliminary data set, MR was more sensitive than CT, CT was more sensitive than NM, there were three false negatives, and there were no false positives. Performing two types of cisternogram increased the chance of identifying leaks. More data is needed to draw stronger conclusions in this ongoing study.&#x0D; Conclusion/Potential Impact: The initial data suggests that MR cisternogram with intrathecal gadolinium is superior to CT and SPECT-CT cisternogram for detecting skull base CSF leaks. Localizing skull base CSF leaks helps enable surgeons to perform less invasive corrective procedures.

Peculiar characteristics of ground motions in Southern Italy: Insights from global and regional ground motion models

We investigate peculiar characteristics of ground motions in Southern Italy (e.g. apparent fast anelastic attenuation and trends of event terms at different periods) using a comprehensive dataset of earthquake recordings between 1969 and 2020. By doing so, we gained insights into the relative performance of eight selected region-specific, global, and global with regional adjustment ground motion models (GMMs). Our analysis is performed using a preliminary dataset (i.e. including all ground motions recorded in the area for the selected analysis period) and an independent dataset (i.e. comprising data not used to develop the models). We analyze total residuals, event terms, within-even residuals, and residuals standardized by model standard deviations (i.e. epsilon). The latter is performed to obtain a robust comparison of GMMs with different standard deviation types and levels. These approaches are employed to ground motion characterization studies for the first time in this region. Our results show that in Southern Italy, there is an apparent anelastic attenuation of the ground motion faster than in other seismic districts. Overall, regional models capture this feature better than global models. Regional adjustments to global models better capture the observed anelastic attenuation at large distances. Using the standardized residuals analysis, we observe that all selected GMMs systematically underestimate the observed ground motion for relatively high ground motion levels and its variability at any intensity levels in the study region. These outcomes may help improving future ground motion models and related engineering applications involving such models in performance-based frameworks.

A Real-Time Novelty Recognition Framework Based on Machine Learning for Fault Detection

New technologies are developed inside today’s companies with the ascent of Industry 4.0 paradigm; Artificial Intelligence applied to Predictive Maintenance is one of these, helping factories automate their systems in detecting anomalies. The deviation of statistical features from standard operating conditions computed on collected data is a common investigation technique that companies use. The information loss due to transformation from raw data to extracted features is a problem of this approach. Furthermore, a common Predictive Maintenance framework requires historical data about failures that often do not exist, neglecting the possibility of applying it. This paper uses Artificial Intelligence as Machine Learning models to recognize when something changes in the data’s behavior collected up to that moment, also helping companies to gather a preliminary dataset for future Predictive Maintenance implementation. The aim concerns a framework in which several sensors are used to collect data by adopting a sensor fusion approach. The architecture is composed of an optimized software system able to enhance the computation scalability and the response time regarding novelty detection. This article analyzes the proposed architecture, then explains a proof-of-concept development using a digital model; finally, two real cases are studied to show how the framework behaves in a real environment. The analysis done in this paper has an application-oriented approach; hence a company can directly use the framework in its systems.

Analysis of a Cardiac-Necrosis-Biomarker Release in Patients with Acute Myocardial Infarction via Nonlinear Mixed-Effects Models

The release of the cardiac troponin T (cTnT) in patients with acute myocardial infarction (AMI) has been analyzed through a methodology based on nonlinear mixed-effects (NME) models. The aim of this work concerns the investigation of any possible relationship between clinical covariates and the dynamics of the release of cTnT to derive more detailed and useful clinical information for the correct treatment of these patients. An ad-hoc mechanistic model describing the biomarker release process after AMI has been devised, assessed, and exploited to evaluate the impact of the available clinical covariates on the cTnT release dynamic. The following approach was tested on a preliminary dataset composed of a small number of potential clinical covariates: employing an unsupervised approach, and despite the limited sample size, dyslipidemia, a known risk factor for cardiovascular disease, was found to be a statistically significant covariate. By increasing the number of covariates considered in the model, and patient cohort, we envisage that this approach may provide an effective means to automatically classify AMI patients and to investigate the role of interactions between clinical covariates and cTnT release.

Construction of a Character Dataset for Historical Uchen Tibetan Documents under Low-Resource Conditions

The construction of a character dataset is an important part of the research on document analysis and recognition of historical Tibetan documents. The results of character segmentation research in the previous stage are presented by coloring the characters with different color values. On this basis, the characters are annotated, and the character images corresponding to the annotation are extracted to construct a character dataset. The construction of a character dataset is carried out as follows: (1) text annotation of segmented characters is performed; (2) the character image is extracted from the character block based on the real position information; (3) according to the class of annotated text, the extracted character images are classified to construct a preliminary character dataset; (4) data augmentation is used to solve the imbalance of classes and samples in the preliminary dataset; (5) research on character recognition based on the constructed dataset is performed. The experimental results show that under low-resource conditions, this paper solves the challenges in the construction of a historical Uchen Tibetan document character dataset and constructs a 610-class character dataset. This dataset lays the foundation for the character recognition of historical Tibetan documents and provides a reference for the construction of relevant document datasets.

Abstract 11930: Ascension Health System Experience With Extended Ambulatory Electrocardiographic Monitoring: Implementation of a Novel Integrated Approach to Better Characterize Disparities in Healthcare

Introduction: Atrial fibrillation (AF) is an established risk factor for stroke, hospitalization and mortality. Whether AF differs by race has not been well characterized in the real world setting due to inherent challenges in linking large datasets across multiple databases. Methods: Extended ambulatory electrocardiographic (AECG) monitoring with Zio XT allows for continuous recording of cardiac rhythm data for up to 14 days in indicated patients. A preliminary dataset from the Ascension Health System electronic health records (EHR) archive on patients prescribed the Zio XT patch was retrospectively analyzed. The study population consisted of 6,293 patients treated at 34 Ascension centers in 6 states between January 2020 and April 2022. Patient EHR data were linked to proprietary device-specific rhythm reports for analyses. Results: Among 6,293 patients in whom EHR data were matched to Zio XT findings, the median (IQR) age was 63 (47, 73) years, 3,816 (60.6%) were female, and 1,038 (16.5%) were non-Caucasian. Median device wear time was 7.2 (5.0, 13.9) days. AF was detected in 749 (11.9%) patients, including 326 (5.2%) with 100% AF. Median AF burden was 37.0% (3.4%, 100%). Older white males had the highest risk of AF events, whereas younger non-white females had the lowest risk (Figure 1). Non-white subgroups demonstrated greater freedom from AF compared to whites. Conclusions: In these preliminary findings, the use of Zio XT for extended AECG monitoring in the Ascension Health System is in predominantly older patients. Patients were racially and geographically diverse, and females were well represented. More than 10% of patients had AF detected during the monitoring period. Despite small sample sizes, differences in AF detection by subgroups demonstrate that a tailored approach to managing AF remains a critically important research question. Linkage of patient EHR data with device-specific outcomes can be a useful approach to address this important unmet need.

H2, CO2, and CH4 Adsorption Potential of Kerogen as a Function of Pressure, Temperature, and Maturity.

We performed molecular dynamics simulation to elucidate the adsorption behavior of hydrogen (H2), carbon dioxide (CO2), and methane (CH4) on four sub-models of type II kerogens (organic matter) of varying thermal maturities over a wide range of pressures (2.75 to 20 MPa) and temperatures (323 to 423 K). The adsorption capacity was directly correlated with pressure but indirectly correlated with temperature, regardless of the kerogen or gas type. The maximum adsorption capacity was 10.6 mmol/g for the CO2, 7.5 mmol/g for CH4, and 3.7 mmol/g for the H2 in overmature kerogen at 20 MPa and 323 K. In all kerogens, adsorption followed the trend CO2 &gt; CH4 &gt; H2 attributed to the larger molecular size of CO2, which increased its affinity toward the kerogen. In addition, the adsorption capacity was directly associated with maturity and carbon content. This behavior can be attributed to a specific functional group, i.e., H, O, N, or S, and an increase in the effective pore volume, as both are correlated with organic matter maturity, which is directly proportional to the adsorption capacity. With the increase in carbon content from 40% to 80%, the adsorption capacity increased from 2.4 to 3.0 mmol/g for H2, 7.7 to 9.5 mmol/g for CO2, and 4.7 to 6.3 mmol/g for CH4 at 15 MPa and 323 K. With the increase in micropores, the porosity increased, and thus II-D offered the maximum adsorption capacity and the minimum II-A kerogen. For example, at a fixed pressure (20 MPa) and temperature (373 K), the CO2 adsorption capacity for type II-A kerogen was 7.3 mmol/g, while type II-D adsorbed 8.9 mmol/g at the same conditions. Kerogen porosity and the respective adsorption capacities of all gases followed the order II-D &gt; II-C &gt; II-B &gt; II-A, suggesting a direct correlation between the adsorption capacity and kerogen porosity. These findings thus serve as a preliminary dataset on the gas adsorption affinity of the organic-rich shale reservoirs and have potential implications for CO2 and H2 storage in organic-rich formations.

The study on health of older people in Germany (Health 65+): Design and first results

Abstract Background Integrating older persons into continuous national public health monitoring is crucial but challenging. Health 65+ is the first national health interview and examination survey in Germany specifically tailored to the needs of the population 65+ years. Methods Health 65+ is based on two-stage stratified random sampling from 128 local population registries. It is based on a previously tested three-step procedure to contact the study population (letter, telephone, home visits). 12,448 individuals 65+ years were invited for survey participation between June 2021 and April 2022. Participation comprised answering a baseline-questionnaire/interview covering a consented set of key health indicators including SARS-COV2 infections, and 3 follow-ups. After one year participants are invited to an examination (e.g. blood pressure, grip strength, cognitive function) during a home visit. All-cause mortality, health insurance data and information on social and built environment will be linked to survey data. The preliminary data set comprises 3,107 baseline participants. Preliminary results Only few individuals were excluded for survey participation as they had deceased before invitation, moved to an unknown residence or had insufficient German language skills. Of the adjusted gross sample, 32% took part in the survey (47.9% women, mean age 78.8 years). Only 3.3% of the participants did not receive any vaccination against COVID-19, with no differences between gender or age-groups. 3.5% had already experienced a laboratory-confirmed SARS-CoV-2 infection. Conclusions Health 65+ collects information that cannot be obtained from any other data sources. In combination with information from routine health data and official health statistics, the results will aid health policy planning and implementation research to improve health and wellbeing of older people in Germany. For example, preliminary results show, that vaccination acceptance was high in persons 65+ years in Germany. Key messages • Health 65+ will provide health data of people 65+ years in Germany. • The results will aid health policy planning.

The preliminary DNA barcoding dataset of fishes from Chinese coastal waters

DNA barcoding is a molecular method for species identification based on a short standardized DNA sequence derived from the 5’ region of the cytochrome c oxidase subunit I (COI) gene, which is widely applied in marine fish identification. In this study, DNA barcodes were generated for 1,671 specimens representing 309 morphological marine fish species belonging to 223 genera and covering 101 families from Chinese coastal waters. The mean Kimura 2-parameter (K2P) genetic distances within species, genera, families, and orders were 0.20%, 20.94%, 24.45%, and 25.34%, respectively. As expected, the mean interspecific distance was higher than the mean intraspecific distance. Phylogenetic trees based on Bayesian inference (BI) and neighbor-joining (NJ) methods agreed with the taxonomic classifications. A total of 306 operational taxonomic units (OTUs) were reidentified via automatic barcode gap discovery (ABGD) analysis. An estimated nucleotide substitution matrices showed a transversion bias compared to transitional mutation. The amino acid variations of the barcoding region were also examined, and 13 amino acid sites exhibited high variability. Overall, our findings confirmed that DNA barcoding is a highly reliable mean of identifying fish species in Chinese coastal waters, and that the barcode library established here would facilitate the monitoring and management of the marine fishery resources in China.

Stabilizing effects of group formation by Serengeti herbivores on predator-prey dynamics

Predator-prey theory often assumes that potential prey individuals are solitary and evenly distributed in space. This assumption is violated in social, mobile prey, such as many ungulates. Here we use data from 80 monthly field censuses to estimate the parameters for a power relationship between herd density and population density for eight species of large herbivores commonly found in the diet of Serengeti lions, confirming a power relationship proposed from a preliminary Serengeti dataset. Here we extend our analysis of that model to demonstrate how parameters of the power function relate to average herd size and density-dependent changes in herd size and evaluate how interspecific variation in these parameters shapes the group-dependent functional response by Serengeti lions for eight prey species. We apply the different prey-specific functional response models in a Rosenzweig-MacArthur framework to compare their impact on the stability of predator–prey dynamics. Model outcomes suggest that group formation plays a strong role in stabilizing lion–herbivore interactions in Serengeti by forcing lions to search over a larger area before each prey encounter. As a consequence of grouping by their prey, our model also suggests that Serengeti lions are forced to broaden their diets to include multiple species of prey in order to persist, potentially explaining the generalist foraging by lions routinely recorded across multiple ecosystems.

Hierarchical optimal designs and modeling for engineering: A case‐study in the rail sector

AbstractComplex engineering and technological processes typically generate data with a non‐trivial hierarchical structure. To this end, in this article we propose a full procedure for optimizing such processes through optimal experimental designs and modeling. In order to study a hierarchical structure, several types of experimental factors may arise, making the building of the experimental design challenging. Starting from the analysis of a preliminary dataset and a pilot design including nested, branching, and shared experimental factors, as well as a new type of experimental factor called composite‐form‐factor, we build a hierarchical D‐optimal experimental design using genetic algorithms. We apply our proposal to a real case‐study in the rail sector aimed at optimizing the payload distribution of freight trains. In this case‐study we also achieve the best train configuration by minimizing the in‐train forces. The results are very satisfactory and confirm that our full procedure represents a valid method to be successfully applied for solving similar technological problems.

AEYE: A deep learning system for video nystagmus detection.

BackgroundNystagmus identification and interpretation is challenging for non-experts who lack specific training in neuro-ophthalmology or neuro-otology. This challenge is magnified when the task is performed via telemedicine. Deep learning models have not been heavily studied in video-based eye movement detection.MethodsWe developed, trained, and validated a deep-learning system (aEYE) to classify video recordings as normal or bearing at least two consecutive beats of nystagmus. The videos were retrospectively collected from a subset of the monocular (right eye) video-oculography (VOG) recording used in the Acute Video-oculography for Vertigo in Emergency Rooms for Rapid Triage (AVERT) clinical trial (#NCT02483429). Our model was derived from a preliminary dataset representing about 10% of the total AVERT videos (n = 435). The videos were trimmed into 10-sec clips sampled at 60 Hz with a resolution of 240 × 320 pixels. We then created 8 variations of the videos by altering the sampling rates (i.e., 30 Hz and 15 Hz) and image resolution (i.e., 60 × 80 pixels and 15 × 20 pixels). The dataset was labeled as “nystagmus” or “no nystagmus” by one expert provider. We then used a filtered image-based motion classification approach to develop aEYE. The model's performance at detecting nystagmus was calculated by using the area under the receiver-operating characteristic curve (AUROC), sensitivity, specificity, and accuracy.ResultsAn ensemble between the ResNet-soft voting and the VGG-hard voting models had the best performing metrics. The AUROC, sensitivity, specificity, and accuracy were 0.86, 88.4, 74.2, and 82.7%, respectively. Our validated folds had an average AUROC, sensitivity, specificity, and accuracy of 0.86, 80.3, 80.9, and 80.4%, respectively. Models created from the compressed videos decreased in accuracy as image sampling rate decreased from 60 Hz to 15 Hz. There was only minimal change in the accuracy of nystagmus detection when decreasing image resolution and keeping sampling rate constant.ConclusionDeep learning is useful in detecting nystagmus in 60 Hz video recordings as well as videos with lower image resolutions and sampling rates, making it a potentially useful tool to aid future automated eye-movement enabled neurologic diagnosis.

Fast intraoperative histology-based diagnosis of gliomas with third harmonic generation microscopy and deep learning

Management of gliomas requires an invasive treatment strategy, including extensive surgical resection. The objective of the neurosurgeon is to maximize tumor removal while preserving healthy brain tissue. However, the lack of a clear tumor boundary hampers the neurosurgeon’s ability to accurately detect and resect infiltrating tumor tissue. Nonlinear multiphoton microscopy, in particular higher harmonic generation, enables label-free imaging of excised brain tissue, revealing histological hallmarks within seconds. Here, we demonstrate a real-time deep learning-based pipeline for automated glioma image analysis, matching video-rate image acquisition. We used a custom noise detection scheme, and a fully-convolutional classification network, to achieve on average 79% binary accuracy, 0.77 AUC and 0.83 mean average precision compared to the consensus of three pathologists, on a preliminary dataset. We conclude that the combination of real-time imaging and image analysis shows great potential for intraoperative assessment of brain tissue during tumor surgery.

Introducing the Digital Humanities in Ireland Landscape Report Dataset

The UK-Ireland Digital Humanities Network was jointly funded in July 2020 by the Irish Research Council (IRC) and the UK Research and Innovation’s (UKRI) Arts and Humanities Research Council (AHRC) under the ground-breaking Collaboration in Digital Humanities Networking Grant Scheme. The joint aims of the Network were to: a) undertake research and consultation towards the implementation of a permanent UK-Ireland Digital Humanities Association; and b) to develop a clear roadmap for collaboration in the field between the two countries. An ancillary objective of the Irish Network members is to provide an up-to-date evaluation of the role and scope of Digital Humanities in Ireland, both past and present, to facilitate longer-term thinking about Digital Humanities so that we might optimise future developments in the field, including the nascent UK-Ireland Digital Humanities Association. To that end, the respective partners are developing a Digital Humanities in Ireland Landscape Report. The research informing the Landscape Report will be delivered in two phases. The initial phase took place between March and September 2021 and comprised the identification via desk research, collection and collation of data pertaining to Digital Humanities entities in Ireland. The second phase of the data gathering/collection exercise entails the presentation of the preliminary dataset to the wider Digital Humanities community for input and suggestions. To that end, we have created an Open Science Framework (OSF) repository. This contribution introduces the Digital Humanities in Ireland Landscape Report dataset, its methodology and primary sources and offers some preliminary observations and analysis. It concludes with some suggestions for potential use cases and further directions for the dataset.