A Multi-attention Collaborative Deep Learning Approach for Blood Pressure Prediction

We develop a deep learning model based on Long Short-term Memory (LSTM) to predict blood pressure based on a unique data set collected from physical examination centers capturing comprehensive multi-year physical examination and lab results. In the Multi-attention Collaborative Deep Learning model (MAC-LSTM) we developed for this type of data, we incorporate three types of attention to generate more explainable and accurate results. In addition, we leverage information from similar users to enhance the predictive power of the model due to the challenges with short examination history. Our model significantly reduces predictive errors compared to several state-of-the-art baseline models. Experimental results not only demonstrate our model’s superiority but also provide us with new insights about factors influencing blood pressure. Our data is collected in a natural setting instead of a setting designed specifically to study blood pressure, and the physical examination items used to predict blood pressure are common items included in regular physical examinations for all the users. Therefore, our blood pressure prediction results can be easily used in an alert system for patients and doctors to plan prevention or intervention. The same approach can be used to predict other health-related indexes such as BMI.

Low-Altitude Aerial Video Surveillance via One-Class SVM Anomaly Detection from Textural Features in UAV Images

In recent years, small-scale Unmanned Aerial Vehicles (UAVs) have been used in many video surveillance applications, such as vehicle tracking, border control, dangerous object detection, and many others. Anomaly detection can represent a prerequisite of many of these applications thanks to its ability to identify areas and/or objects of interest without knowing them a priori. In this paper, a One-Class Support Vector Machine (OC-SVM) anomaly detector based on customized Haralick textural features for aerial video surveillance at low-altitude is presented. The use of a One-Class SVM, which is notoriously a lightweight and fast classifier, enables the implementation of real-time systems even when these are embedded in low-computational small-scale UAVs. At the same time, the use of textural features allows a vision-based system to detect micro and macro structures of an analyzed surface, thus allowing the identification of small and large anomalies, respectively. The latter aspect plays a key role in aerial video surveillance at low-altitude, i.e., 6 to 15 m, where the detection of common items, e.g., cars, is as important as the detection of little and undefined objects, e.g., Improvised Explosive Devices (IEDs). Experiments obtained on the UAV Mosaicking and Change Detection (UMCD) dataset show the effectiveness of the proposed system in terms of accuracy, precision, recall, and F1-score, where the model achieves a 100% precision, i.e., never misses an anomaly, but at the expense of a reasonable trade-off in its recall, which still manages to reach up to a 71.23% score. Moreover, when compared to classical Haralick textural features, the model obtains significantly higher performances, i.e., ≈20% on all metrics, further demonstrating the approach effectiveness.

Patients’ Perspectives on Qualitative Olfactory Dysfunction: Thematic Analysis of Social Media Posts

Background The impact of qualitative olfactory disorders is underestimated. Parosmia, the distorted perception of familiar odors, and phantosmia, the experience of odors in the absence of a stimulus, can arise following postinfectious anosmia, and the incidences of both have increased substantially since the outbreak of COVID-19. Objective The aims of this study are to explore the symptoms and sequalae of postinfectious olfactory dysfunction syndrome using unstructured and unsolicited threads from social media, and to articulate the perspectives and concerns of patients affected by these debilitating olfactory disorders. Methods A thematic analysis and content analysis of posts in the AbScent Parosmia and Phantosmia Support group on Facebook was conducted between June and December 2020. Results In this paper, we identify a novel symptom, olfactory perseveration, which is a triggered, identifiable, and usually unpleasant olfactory percept that persists in the absence of an ongoing stimulus. We also observe fluctuations in the intensity and duration of symptoms of parosmia, phantosmia, and olfactory perseveration. In addition, we identify a group of the most common items (coffee, meat, onion, and toothpaste) that trigger distortions; however, people have difficulty describing these distortions, using words associated with disgust and revulsion. The emotional aspect of living with qualitative olfactory dysfunction was evident and highlighted the detrimental impact on mental health. Conclusions Qualitative and unsolicited data acquired from social media has provided useful insights into the patient experience of parosmia and phantosmia, which can inform rehabilitation strategies and ongoing research into understanding the molecular triggers associated with parosmic distortions and research into patient benefit.

Pre-Statistical Harmonization of Behavioral Instruments Across Eight Surveys and Trials

Data harmonization methods facilitate further use of existing studies and research resources. Most statistical harmonization methods require pooling data across studies, which is complex and requires careful scrutiny of source data. Most methods (e.g., item response theory) require datasets to have common items for linking a common construct across studies: this necessitates the qualitative process of pre-statistical harmonization. Here, we document pre-statistical harmonization of items measuring behavioral and psychological symptoms (e.g., agitation, wandering, etc.) which represent problematic behaviors among people with dementia administered in a national survey (ADAMS), evaluations conducted at Alzheimer’s Disease Research Centers (NACC), and in six randomized trials (COPE, TAP, ALZQOL, ACT, REACH, ADSPlus). We describe our approach to review question content and scoring procedures to establish comparability across items prior to data pooling. We identified 327 items from 15 instruments across these eight studies. We found considerable cross-study heterogeneity in administration and coding procedures for items that measure the same domain. For example, eight items were coded as count variables in some studies but as categorical variables in others. Moreover, of the 359 items, 191 are conditionally dependent on values of another item. These issues around item response heterogeneity and conditional dependency needed to be resolved prior to estimation of item response theory models for statistical co-calibration. We leveraged several rigorous data transformation procedures to address these issues, including re-coding and winsorization. This study provides guidelines for how future research may acknowledge and address similar issues in pooling behavioral and related instruments.

A CATEGORY-THEORY APPROACH FOR CONSTRUCTION ONTOLOGIES IN SUBSURFACE MASS TRANSIT

The construction and expansion of subway systems represents an important step towards better livability conditions in a rapidly urbanizing world. However, underground construction has not benefited from well-established ontologies of semantic and geometric representation, such as Building Information Modelling (which is used for standalone structures) and City Geography Markup Language (which is designed for continuous urban elements). To bridge that gap, this paper proposes a novel and highly flexible means to underpin a relevant ontology. The approach uses the ontology log, or olog, a model of knowledge representation based on Category Theory. In an olog, dependencies between objects are restricted to functional relationships (for every object there is a unique correspondence). This robust mathematical formulation allows for a more flexible, yet also informative and user-readable model of the studied entities. In this paper, the olog’s usability is demonstrated through the ontological representation of common items in the fare-control areas of two New York City metro stations. Ologs are shown to capture similar underlying structures both across different stations and within the same station. Importantly, the olog allows for further generalization to incorporate pre-existing data, as well as being a transferable framework for conceptualizations of other metro systems.

Wolf Use of Humanmade Objects During Pup-rearing

Some animals use humanmade objects for building and constructing nests or shelter and even for play. Gray wolves (Canis lupus) gather and use humanmade objects discovered in their natural environment. Gathering humanmade objects is a peculiar behavior particularly when there is no immediately apparent benefit to survival or reproduction. I opportunistically documented 46 different types of humanmade objects with plastic bottles and aluminum cans being the most common items found at wolf pup-rearing sites. Many objects were made of materials that appeared suitable to alleviate pain in teething pups. For some objects, however, it was not immediately obvious that they would alleviate teething pain due to their unpliable material. Additionally, such objects were quite rare in wolves’ natural environment although it was not uncommon to find them at pup-rearing sites. Rare humanmade objects may provide a novelty that stimulates pups more than common objects. I hypothesize that objects used by wolf pups 1) alleviate pain from teething, and 2) provide adults respite from energetic pups. The latter is an important distinction because it implies the benefit of object play is to the adults and not the pups per se. Gathering novel objects that occupy energetic and hungry pups may influence the overall ability of social carnivores to leave young unattended while they hunt, to rest upon their return, and ultimately rear young successfully.