Harnessing AI and robotics in humanitarian assistance and disaster response.

AI and robotics can facilitate humanitarian assistance and disaster response, but partnerships with practitioners are crucial.

Deep learning models for COVID-19 chest x-ray classification: Preventing shortcut learning using feature disentanglement.

In response to the COVID-19 global pandemic, recent research has proposed creating deep learning based models that use chest radiographs (CXRs) in a variety of clinical tasks to help manage the crisis. However, the size of existing datasets of CXRs from COVID-19+ patients are relatively small, and researchers often pool CXR data from multiple sources, for example, using different x-ray machines in various patient populations under different clinical scenarios. Deep learning models trained on such datasets have been shown to overfit to erroneous features instead of learning pulmonary characteristics in a phenomenon known as shortcut learning. We propose adding feature disentanglement to the training process. This technique forces the models to identify pulmonary features from the images and penalizes them for learning features that can discriminate between the original datasets that the images come from. We find that models trained in this way indeed have better generalization performance on unseen data; in the best case we found that it improved AUC by 0.13 on held out data. We further find that this outperforms masking out non-lung parts of the CXRs and performing histogram equalization, both of which are recently proposed methods for removing biases in CXR datasets.

An Artificial Intelligence Dataset for Solar Energy Locations in India.

Rapid development of renewable energy sources, particularly solar photovoltaics (PV), is critical to mitigate climate change. As a result, India has set ambitious goals to install 500 gigawatts of solar energy capacity by 2030. Given the large footprint projected to meet renewables energy targets, the potential for land use conflicts over environmental values is high. To expedite development of solar energy, land use planners will need access to up-to-date and accurate geo-spatial information of PV infrastructure. In this work, we developed a spatially explicit machine learning model to map utility-scale solar projects across India using freely available satellite imagery with a mean accuracy of 92%. Our model predictions were validated by human experts to obtain a dataset of 1363 solar PV farms. Using this dataset, we measure the solar footprint across India and quantified the degree of landcover modification associated with the development of PV infrastructure. Our analysis indicates that over 74% of solar development In India was built on landcover types that have natural ecosystem preservation, or agricultural value.

Effective deep learning approaches for predicting COVID-19 outcomes from chest computed tomography volumes.

The rapid evolution of the novel coronavirus disease (COVID-19) pandemic has resulted in an urgent need for effective clinical tools to reduce transmission and manage severe illness. Numerous teams are quickly developing artificial intelligence approaches to these problems, including using deep learning to predict COVID-19 diagnosis and prognosis from chest computed tomography (CT) imaging data. In this work, we assess the value of aggregated chest CT data for COVID-19 prognosis compared to clinical metadata alone. We develop a novel patient-level algorithm to aggregate the chest CT volume into a 2D representation that can be easily integrated with clinical metadata to distinguish COVID-19 pneumonia from chest CT volumes from healthy participants and participants with other viral pneumonia. Furthermore, we present a multitask model for joint segmentation of different classes of pulmonary lesions present in COVID-19 infected lungs that can outperform individual segmentation models for each task. We directly compare this multitask segmentation approach to combining feature-agnostic volumetric CT classification feature maps with clinical metadata for predicting mortality. We show that the combination of features derived from the chest CT volumes improve the AUC performance to 0.80 from the 0.52 obtained by using patients' clinical data alone. These approaches enable the automated extraction of clinically relevant features from chest CT volumes for risk stratification of COVID-19 patients.

Gelatinisation and milling whole-wheat increases postprandial blood glucose: randomised crossover study of adults with type 2 diabetes.

AIMS/HYPOTHESES: We examined the effects of milling and cooking whole grains in water to achieve starch gelatinisation on postprandial blood glucose using a randomised crossover open-label design. Participants were adults with type 2 diabetes whose body weight or medications had not changed in at least 3 months. METHODS: Postprandial blood glucose (measured as incremental AUC [iAUC]) was measured following consumption of four nutrient-matched whole-wheat porridge test-meals. Test-meals included gelatinised or native starch and were made with either finely milled or intact whole-wheat. RESULTS: Eighteen adults (63.1 ± 9.8 years, HbA1c 57.0 ± 11.5 mmol/mol [7.4 ± 3.2%]) completed the study. iAUC was higher following cooked meals (gelatinised starch) than following uncooked meals (native starch) (mean difference [MD] 268, 95% CI 188, 348 mmol/l × min). Consuming finely milled whole-wheat produced a higher iAUC compared with intact whole-wheat (MD 173, 95% CI 80, 266 mmol/l × min). There was no evidence of an interaction effect (p = 0.841). CONCLUSIONS: Both the nature of starch and the grain structure of whole-wheat influence the glycaemic response of adults with type 2 diabetes mellitus. FUNDING: Baking Industry Research Trust of New Zealand and the Riddet Centre of Research Excellence. TRIAL REGISTRATION: www.anzctr.org.au ACTRN12617000328370.

Deep learning models for COVID-19 chest x-ray classification: Preventing shortcut learning using feature disentanglement.

In response to the COVID-19 global pandemic, recent research has proposed creating deep learning based models that use chest radiographs (CXRs) in a variety of clinical tasks to help manage the crisis. However, the size of existing datasets of CXRs from COVID-19+ patients are relatively small, and researchers often pool CXR data from multiple sources, for example, using different x-ray machines in various patient populations under different clinical scenarios. Deep learning models trained on such datasets have been shown to overfit to erroneous features instead of learning pulmonary characteristics -- a phenomenon known as shortcut learning. We propose adding feature disentanglement to the training process, forcing the models to identify pulmonary features from the images while penalizing them for learning features that can discriminate between the original datasets that the images come from. We find that models trained in this way indeed have better generalization performance on unseen data; in the best case we found that it improved AUC by 0.13 on held out data. We further find that this outperforms masking out non-lung parts of the CXRs and performing histogram equalization, both of which are recently proposed methods for removing biases in CXR datasets.

Emoji can facilitate recognition of conveyed indirect meaning.

In face-to-face communication there are multiple paralinguistic and gestural features that facilitate recognition of a speaker's intended meaning, features that are lacking when people communicate digitally (e.g., texting). As a result, substitutes have emerged (expressive punctuation, capitalization, etc.) to facilitate communication in these situations. However, little is known about the comprehension processes involved in digital communication. In this research we examined the role of emoji in the comprehension of face-threatening, indirect replies. Participants in two experiments read question-reply sequences and then judged the accuracy of interpretations of the replies. On critical trials the reply violated the relation maxim and conveyed a negative, face-threatening response. On one-third of the trials the reply contained only text, on one-third of the trials the reply contained text and an emoji, and on one-third of the trials the reply contained only an emoji. When the question requested potentially negative information about one of the interactants (disclosures and opinions), participants were more likely to endorse the indirect meaning of the reply, and did so faster, when the reply contained an emoji than when it did not. This effect did not occur when the question was a request for action, a more conventional type of indirect reply. Overall, then, this research demonstrates that emoji can sometimes facilitate the comprehension of meaning. Future research is needed to examine the boundary conditions for this effect.

Create Our Own Kai: A Randomised Control Trial of a Cooking Intervention with Group Interview Insights into Adolescent Cooking Behaviours.

Cooking is frequently associated with a healthier diet, however few youth cooking intervention studies have used control groups or follow-ups. Additionally, although cooking is associated with better mental well-being among adolescents, this has not been examined experimentally. This randomised controlled trial investigated whether a five-day intensive holiday cooking program, followed by six weeks of weekly meal kits with Facebook support groups, affected the cooking-related outcomes, diet quality and mental well-being among adolescents, with a 12-month follow-up. Adolescents aged 12-15 years (intervention: n = 91, 60% female; control: n = 27, 78% female) completed baseline, post-intervention and 12-month follow-up anthropometric measures, and questionnaire measures of mental well-being, diet quality and cooking attitudes, self-efficacy and behaviours. The intervention group's post-intervention outcomes improved significantly more for mental well-being, diet quality, helping make dinner, cooking self-efficacy and positive cooking attitude, however body mass index (BMI) z-scores also increased. Differences were maintained at 12 months for self-efficacy only. Group interviews showed that participants' cooking behaviours were strongly influenced by family factors. Adolescent cooking interventions may have many short-term benefits, however cooking self-efficacy appears most responsive and stable over time. Effects on BMI need further investigation. Family factors influence whether and what adolescents cook post-intervention.

Modeling migration patterns in the USA under sea level rise.

Sea level rise in the United States will lead to large scale migration in the future. We propose a framework to examine future climate migration patterns using models of human migration. Our framework requires that we distinguish between historical versus climate driven migration and recognizes how the impacts of climate change can extend beyond the affected area. We apply our framework to simulate how migration, driven by sea level rise, differs from baseline migration patterns. Specifically, we couple a sea level rise model with a data-driven model of human migration and future population projections, creating a generalized joint model of climate driven migration that can be used to simulate population distributions under potential future sea level rise scenarios. The results of our case study suggest that the effects of sea level rise are pervasive, expanding beyond coastal areas via increased migration, and disproportionately affecting some areas of the United States.

Summit: Scaling Deep Learning Interpretability by Visualizing Activation and Attribution Summarizations.

Deep learning is increasingly used in decision-making tasks. However, understanding how neural networks produce final predictions remains a fundamental challenge. Existing work on interpreting neural network predictions for images often focuses on explaining predictions for single images or neurons. As predictions are often computed from millions of weights that are optimized over millions of images, such explanations can easily miss a bigger picture. We present Summit, an interactive system that scalably and systematically summarizes and visualizes what features a deep learning model has learned and how those features interact to make predictions. Summit introduces two new scalable summarization techniques: (1) activation aggregation discovers important neurons, and (2) neuron-influence aggregation identifies relationships among such neurons. Summit combines these techniques to create the novel attribution graph that reveals and summarizes crucial neuron associations and substructures that contribute to a model's outcomes. Summit scales to large data, such as the ImageNet dataset with 1.2M images, and leverages neural network feature visualization and dataset examples to help users distill large, complex neural network models into compact, interactive visualizations. We present neural network exploration scenarios where Summit helps us discover multiple surprising insights into a prevalent, large-scale image classifier's learned representations and informs future neural network architecture design. The Summit visualization runs in modern web browsers and is open-sourced.

Wholegrain Particle Size Influences Postprandial Glycemia in Type 2 Diabetes: A Randomized Crossover Study Comparing Four Wholegrain Breads.

OBJECTIVE: Wholegrain foods vary in the extent of processing. We investigated whether wholegrain particle size in bread influences postprandial glycemia in type 2 diabetes. RESEARCH DESIGN AND METHODS: Postprandial glycemia (incremental area under the blood glucose curve [iAUC]) was measured after consumption of three breads made with roller-milled wholegrain flour and added grains and a fourth made with stone-ground flour. All flours and grains were 100% wholegrain wheat. Breads were nutrient matched. RESULTS: Fifteen adults (64 ± 10 years, HbA1c 58 ± 13 mmol/mol) completed the study. iAUC for the three breads made with roller-milled flour ranged from 376 to 641 mmol-1min-1, inverse linear trend for grain particle size P = 0.039. The iAUC for stone-ground wholegrain bread (503) was smaller than predicted from mean particle size. CONCLUSIONS: Wholegrain structural integrity in bread is a determinant of glycemic response. These findings have implications for dietary advice and the definition of the term "'wholegrain."

Pilot Testing of an Intensive Cooking Course for New Zealand Adolescents: The Create-Our-Own Kai Study.

The role of cooking on health and wellbeing is a recent area of scientific interest. In order to investigate this role, a cooking program that is suitable for each target population is needed e.g., a program designed for American or Australian children might not be appropriate for teenagers in New Zealand. As there was no similar previously evaluated program already available, the study’s purpose was to test an intensive cooking intervention on cooking confidence and knowledge amongst a group of adolescents from Dunedin, New Zealand, and to assess its acceptability to participants. This five-day program comprised interactive cooking sessions and informal nutrition education and ran from 9 a.m. to 3:30 p.m., Monday to Friday during school holidays. Participants completed questionnaires on cooking skills and confidence at baseline and the end of intervention and took part in a group interview, which aimed to investigate the acceptability and outcome of the program. Twenty-one participants aged between 12 and 16 years old completed the program. At the end of the program, significant increases were seen in both skills and confidence levels, and feedback from the group interview indicated that the participants enjoyed the program and that it provided additional results other than those that were cooking related.

Sex differences in hippocampal damage, cognitive impairment, and trophic factor expression in an animal model of an alcohol use disorder.

Compared to men, women disproportionally experience alcohol-related organ damage, including brain damage, and while men remain more likely to drink and to drink heavily, there is cause for concern because women are beginning to narrow the gender gap in alcohol use disorders. The hippocampus is a brain region that is particularly vulnerable to alcohol damage, due to cell loss and decreased neurogenesis. In the present study, we examined sex differences in hippocampal damage following binge alcohol. Consistent with our prior findings, we found a significant binge-induced decrement in dentate gyrus (DG) granule neurons in the female DG. However, in the present study, we found no significant decrement in granule neurons in the male DG. We show that the decrease in granule neurons in females is associated with both spatial navigation impairments and decreased expression of trophic support molecules. Finally, we show that post-binge exercise is associated with an increase in trophic support and repopulation of the granule neuron layer in the female hippocampus. We conclude that sex differences in alcohol-induced hippocampal damage are due in part to a paucity of trophic support and plasticity-related signaling in females.

Altered discharges of spinal neurons parallel the behavioral phenotype shown by rats with bortezomib related chemotherapy induced peripheral neuropathy.

Bortezomib is a first generation proteasome inhibitor that is the frontline chemotherapy for multiple myeloma with the chief dose-limiting side effect of painful peripheral neuropathy. The goal of this study was to define the behavioral phenotype in a preclinical model of bortezomib chemotherapy-induced peripheral neuropathy (CIPN) and to test whether this is matched by changes in the physiological responses of spinal wide dynamic range neurons. Sprague-Dawley rats were treated with four injections of bortezomib at four doses, 0.05mg/kg, 0.10mg/kg, 0.15mg/kg, 0.20mg/kg, or equal volume of saline. All doses of bortezomib above 0.05mg/kg produced showed significant dose-dependent mechanical hyperalgesia that was fully established at 30 days after treatment and that recovered to baseline levels by day 69 after treatment. Thermal, cold, and motor testing were all unaffected by treatment with bortezomib. Spinal wide dynamic range (WDR) neurons in rats with confirmed bortzomib-related CIPN showed an increase in number of evoked discharges to mechanical stimuli and exaggerated after-discharges in rats with bortezomib CIPN.

An overview of animal models of pain: disease models and outcome measures.

UNLABELLED: Pain is ultimately a perceptual phenomenon. It is built from information gathered by specialized pain receptors in tissue, modified by spinal and supraspinal mechanisms, and integrated into a discrete sensory experience with an emotional valence in the brain. Because of this, studying intact animals allows the multidimensional nature of pain to be examined. A number of animal models have been developed, reflecting observations that pain phenotypes are mediated by distinct mechanisms. Animal models of pain are designed to mimic distinct clinical diseases to better evaluate underlying mechanisms and potential treatments. Outcome measures are designed to measure multiple parts of the pain experience, including reflexive hyperalgesia measures, sensory and affective dimensions of pain, and impact of pain on function and quality of life. In this review, we discuss the common methods used for inducing each of the pain phenotypes related to clinical pain syndromes as well as the main behavioral tests for assessing pain in each model. PERSPECTIVE: Understanding animal models and outcome measures in animals will assist in translating data from basic science to the clinic.

Spinal astrocyte gap junctions contribute to oxaliplatin-induced mechanical hypersensitivity.

UNLABELLED: Spinal glial cells contribute to the development of many types of inflammatory and neuropathic pain. Here the contribution of spinal astrocytes and astrocyte gap junctions to oxaliplatin-induced mechanical hypersensitivity was explored. The expression of glial fibrillary acidic protein (GFAP) in spinal dorsal horn was significantly increased at day 7 but recovered at day 14 after oxaliplatin treatment, suggesting a transient activation of spinal astrocytes by chemotherapy. Astrocyte-specific gap junction protein connexin 43 (Cx43) was significantly increased in dorsal horn at both day 7 and day 14 following chemotherapy, but neuronal (connexin 36 [Cx36]) and oligodendrocyte (connexin 32 [Cx32]) gap junction proteins did not show any change. Blockade of astrocyte gap junction with carbenoxolone (CBX) prevented oxaliplatin-induced mechanical hypersensitivity in a dose-dependent manner and the increase of spinal GFAP expression, but had no effect once the mechanical hypersensitivity induced by oxaliplatin had fully developed. These results suggest that oxaliplatin chemotherapy induces the activation of spinal astrocytes and this is accompanied by increased expression of astrocyte-astrocyte gap junction connections via Cx43. These alterations in spinal astrocytes appear to contribute to the induction but not the maintenance of oxaliplatin-induced mechanical hypersensitivity. Combined, these results suggest that targeting spinal astrocyte/astrocyte-specific gap junction could be a new therapeutic strategy to prevent oxaliplatin-induced neuropathy. PERSPECTIVE: Spinal astrocytes but not microglia were recently shown to be recruited in paclitaxel-related chemoneuropathy. Here, spinal astrocyte gap junctions are shown to play an important role in the induction of oxaliplatin neuropathy.