Elucidating the semantics-topology trade-off for knowledge inference-based pharmacological discovery.

Leveraging AI for synthesizing the deluge of biomedical knowledge has great potential for pharmacological discovery with applications including developing new therapeutics for untreated diseases and repurposing drugs as emergent pandemic treatments. Creating knowledge graph representations of interacting drugs, diseases, genes, and proteins enables discovery via embedding-based ML approaches and link prediction. Previously, it has been shown that these predictive methods are susceptible to biases from network structure, namely that they are driven not by discovering nuanced biological understanding of mechanisms, but based on high-degree hub nodes. In this work, we study the confounding effect of network topology on biological relation semantics by creating an experimental pipeline of knowledge graph semantic and topological perturbations. We show that the drop in drug repurposing performance from ablating meaningful semantics increases by 21% and 38% when mitigating topological bias in two networks. We demonstrate that new methods for representing knowledge and inferring new knowledge must be developed for making use of biomedical semantics for pharmacological innovation, and we suggest fruitful avenues for their development.

Associating biological context with protein-protein interactions through text mining at PubMed scale.

Inferring knowledge from known relationships between drugs, proteins, genes, and diseases has great potential for clinical impact, such as predicting which existing drugs could be repurposed to treat rare diseases. Incorporating key biological context such as cell type or tissue of action into representations of extracted biomedical knowledge is essential for principled pharmacological discovery. Existing global, literature-derived knowledge graphs of interactions between drugs, proteins, genes, and diseases lack this essential information. In this study, we frame the task of associating biological context with protein-protein interactions extracted from text as a classification task using syntactic, semantic, and novel meta-discourse features. We introduce the Insider corpora, which are automatically generated PubMed-scale corpora for training classifiers for the context association task. These corpora are created by searching for precise syntactic cues of cell type and tissue relevancy to extracted regulatory relations. We report F1 scores of 0.955 and 0.862 for identifying relevant cell types and tissues, respectively, for our identified relations. By classifying with this framework, we demonstrate that the problem of context association can be addressed using intuitive, interpretable features. We demonstrate the potential of this approach to enrich text-derived knowledge bases with biological detail by incorporating cell type context into a protein-protein network for dengue fever.

Hybrid Classical-Quantum Transfer Learning for Cardiomegaly Detection in Chest X-rays.

Cardiovascular diseases are among the major health problems that are likely to benefit from promising developments in quantum machine learning for medical imaging. The chest X-ray (CXR), a widely used modality, can reveal cardiomegaly, even when performed primarily for a non-cardiological indication. Based on pre-trained DenseNet-121, we designed hybrid classical-quantum (CQ) transfer learning models to detect cardiomegaly in CXRs. Using Qiskit and PennyLane, we integrated a parameterized quantum circuit into a classic network implemented in PyTorch. We mined the CheXpert public repository to create a balanced dataset with 2436 posteroanterior CXRs from different patients distributed between cardiomegaly and the control. Using k-fold cross-validation, the CQ models were trained using a state vector simulator. The normalized global effective dimension allowed us to compare the trainability in the CQ models run on Qiskit. For prediction, ROC AUC scores up to 0.93 and accuracies up to 0.87 were achieved for several CQ models, rivaling the classical-classical (CC) model used as a reference. A trustworthy Grad-CAM++ heatmap with a hot zone covering the heart was visualized more often with the QC option than that with the CC option (94% vs. 61%, p < 0.001), which may boost the rate of acceptance by health professionals.

Facial expression analysis in a wild sporting environment.

The scientific community and mass media have already reported the use of nonverbal behavior analysis in sports for athletes' performance. Their conclusions stated that certain emotional expressions are linked to athlete's performance, or even that psychological strategies serve to improve endurance performance. This paper examines the portrayal of well-known emotions and their relationship to the participants of an ultra-distance race in a high-stake environment. For this purpose, we analyzed almost 600 runners captured when they passed through a set of locations placed along the race track. We have observed a correlation between the runners' facial expressions and their performance along the track. Moreover, we have analyzed Action Unit activations and aligned our findings with the state-of-the-art psychological baseline.

Contexts and contradictions: a roadmap for computational drug repurposing with knowledge inference.

The cost of drug development continues to rise and may be prohibitive in cases of unmet clinical need, particularly for rare diseases. Artificial intelligence-based methods are promising in their potential to discover new treatment options. The task of drug repurposing hypothesis generation is well-posed as a link prediction problem in a knowledge graph (KG) of interacting of drugs, proteins, genes and disease phenotypes. KGs derived from biomedical literature are semantically rich and up-to-date representations of scientific knowledge. Inference methods on scientific KGs can be confounded by unspecified contexts and contradictions. Extracting context enables incorporation of relevant pharmacokinetic and pharmacodynamic detail, such as tissue specificity of interactions. Contradictions in biomedical KGs may arise when contexts are omitted or due to contradicting research claims. In this review, we describe challenges to creating literature-scale representations of pharmacological knowledge and survey current approaches toward incorporating context and resolving contradictions.

Making Cities Smarter-Optimization Problems for the IoT Enabled Smart City Development: A Mapping of Applications, Objectives, Constraints.

One of the prime aims of smart cities has been to optimally manage the available resources and systems that are used in the city. With an increase in urban population that is set to grow even faster in the future, smart city development has been the main goal for governments worldwide. In this regard, while the useage of Artificial Intelligence (AI) techniques covering the areas of Machine and Deep Learning have garnered much attention for Smart Cities, less attention has focused towards the use of combinatorial optimization schemes. To help with this, the current review presents a coverage of optimization methods and applications from a smart city perspective enabled by the Internet of Things (IoT). A mapping is provided for the most encountered applications of computational optimization within IoT smart cities for five popular optimization methods, ant colony optimization, genetic algorithm, particle swarm optimization, artificial bee colony optimization and differential evolution. For each application identified, the algorithms used, objectives considered, the nature of the formulation and constraints taken in to account have been specified and discussed. Lastly, the data setup used by each covered work is also mentioned and directions for future work have been identified. This review will help researchers by providing them a consolidated starting point for research in the domain of smart city application optimization.

A Deep Convolutional Neural Network-XGB for Direction and Severity Aware Fall Detection and Activity Recognition.

Activity and Fall detection have been a topic of keen interest in the field of ambient assisted living system research. Such systems make use of different sensing mechanisms to monitor human motion and aim to ascertain the activity being performed for health monitoring and other purposes. Towards this end, in addition to activity recognition, fall detection is an especially important task as falls can lead to injuries and sometimes even death. This work presents a fall detection and activity recognition system that not only considers various activities of daily living but also considers detection of falls while taking into consideration the direction and severity. Inertial Measurement Unit (accelerometer and gyroscope) data from the SisFall dataset is first windowed into non-overlapping segments of duration 3 s. After suitable data augmentation, it is then passed on to a Convolutional Neural Network (CNN) for feature extraction with an eXtreme Gradient Boosting (XGB) last stage for classification into the various output classes. The experiments show that the gradient boosted CNN performs better than other comparable techniques, achieving an unweighted average recall of 88%.

Challenges and opportunities in network-based solutions for biological questions.

Network biology is useful for modeling complex biological phenomena; it has attracted attention with the advent of novel graph-based machine learning methods. However, biological applications of network methods often suffer from inadequate follow-up. In this perspective, we discuss obstacles for contemporary network approaches-particularly focusing on challenges representing biological concepts, applying machine learning methods, and interpreting and validating computational findings about biology-in an effort to catalyze actionable biological discovery.

A Hierarchical Approach to Activity Recognition and Fall Detection Using Wavelets and Adaptive Pooling.

Human activity recognition has been a key study topic in the development of cyber physical systems and assisted living applications. In particular, inertial sensor based systems have become increasingly popular because they do not restrict users' movement and are also relatively simple to implement compared to other approaches. In this paper, we present a hierarchical classification framework based on wavelets and adaptive pooling for activity recognition and fall detection predicting fall direction and severity. To accomplish this, windowed segments were extracted from each recording of inertial measurements from the SisFall dataset. A combination of wavelet based feature extraction and adaptive pooling was used before a classification framework was applied to determine the output class. Furthermore, tests were performed to determine the best observation window size and the sensor modality to use. Based on the experiments the best window size was found to be 3 s and the best sensor modality was found to be a combination of accelerometer and gyroscope measurements. These were used to perform activity recognition and fall detection with a resulting weighted F1 score of 94.67%. This framework is novel in terms of the approach to the human activity recognition and fall detection problem as it provides a scheme that is computationally less intensive while providing promising results and therefore can contribute to edge deployment of such systems.

Deep Learning Techniques for Speech Emotion Recognition, from Databases to Models.

The advancements in neural networks and the on-demand need for accurate and near real-time Speech Emotion Recognition (SER) in human-computer interactions make it mandatory to compare available methods and databases in SER to achieve feasible solutions and a firmer understanding of this open-ended problem. The current study reviews deep learning approaches for SER with available datasets, followed by conventional machine learning techniques for speech emotion recognition. Ultimately, we present a multi-aspect comparison between practical neural network approaches in speech emotion recognition. The goal of this study is to provide a survey of the field of discrete speech emotion recognition.

Repurposing biomedical informaticians for COVID-19.

The COVID-19 pandemic is an unprecedented challenge to the biomedical research community at the intersection of great uncertainty due to the novelty of the virus and extremely high stakes due to the large global death count. The global quarantine shut-downs complicated scientific matters because many laboratories were closed down unless they were actively doing COVID-19 related research, making repurposing of activities difficult for many biomedical researchers. Biomedical informaticians, who have been primarily able to continue their research through remote work and video conferencing, have been able to maintain normal activities. In addition to continuing ongoing studies, there has been great grass roots interest in helping in the fight against COVID-19. In this commentary, we describe several projects that arose from this desire to help, and the lessons that the authors learned along the way. We then offer some insights into how these lessons might be applied to make scientific progress be more efficient in future crisis scenarios.

Endoscopic Evaluation of Radiologic Distal Esophageal Thickening.

GOALS: We aim to determine the incidence of esophagogastroduodenoscopies (EGDs) primarily performed for imaging findings of distal esophageal thickening (DET). We also aim to determine if patients with imaging findings of DET have a higher incidence of cancer, and to evaluate the risk factors associated with findings of malignancy. BACKGROUND: The growth of diagnostic imaging has led to an increase in incidental findings of DET. This nonspecific finding frequently prompts an EGD for evaluation-many of which demonstrate benign conditions. There may be a misuse of valuable resources. STUDY: We performed a retrospective chart review of 1080 EGDs from January 2016 to July 2018 at the Veterans Affairs Medical Center, comparing EGDs for the indication of imaging report of DET with EGDs for other indications. Patient demographics, clinical history, imaging, procedure, and pathology reports were collected. Descriptive analysis and biostatistical analysis with χ2, Fisher exact, Wilcoxon rank sum, and Kruskal-Wallis tests were utilized in analyzing the data. RESULTS: Of the 1080 total endoscopies, 8.2% (n=88) were done specifically because of the imaging findings of DET. Those who had EGDs performed because of DET had a higher percentage of abnormal esophageal findings and of cancer. A history of Barrett's esophagus, tobacco use, and having gastrointestinal symptoms were not significant predictors of abnormal findings or of cancer for EGDs done for DET. CONCLUSIONS: There may be a role for EGDs performed for radiologic findings of DET. Even those without risk factors for malignancy should have EGDs performed for DET. Radiologists should consider reporting the DET size in order to determine if significant endoscopic findings correlate with wall thickness.

A topic modeling framework for spatio-temporal information management.

Real-time processing and learning of conflicting data, especially messages coming from different ideas, locations, and time, in a dynamic environment such as Twitter is a challenging task that recently gained lots of attention. This paper introduces a framework for managing, processing, analyzing, detecting, and tracking topics in streaming data. We propose a model selector procedure with a hybrid indicator to tackle the challenge of online topic detection. In this framework, we built an automatic data processing pipeline with two levels of cleaning. Regular and deep cleaning are applied using multiple sources of meta knowledge to enhance data quality. Deep learning and transfer learning techniques are used to classify health-related tweets, with high accuracy and improved F1-Score. In this system, we used visualization to have a better understanding of trending topics. To demonstrate the validity of this framework, we implemented and applied it to health-related twitter data from users originating in the USA over nine months. The results of this implementation show that this framework was able to detect and track the topics at a level comparable to manual annotation. To better explain the emerging and changing topics in various locations over time the result is graphically displayed on top of the United States map.

A Literature-Based Knowledge Graph Embedding Method for Identifying Drug Repurposing Opportunities in Rare Diseases.

Millions of Americans are affected by rare diseases, many of which have poor survival rates. However, the small market size of individual rare diseases, combined with the time and capital requirements of pharmaceutical R&D, have hindered the development of new drugs for these cases. A promising alternative is drug repurposing, whereby existing FDA-approved drugs might be used to treat diseases different from their original indications. In order to generate drug repurposing hypotheses in a systematic and comprehensive fashion, it is essential to integrate information from across the literature of pharmacology, genetics, and pathology. To this end, we leverage a newly developed knowledge graph, the Global Network of Biomedical Relationships (GNBR). GNBR is a large, heterogeneous knowledge graph comprising drug, disease, and gene (or protein) entities linked by a small set of semantic themes derived from the abstracts of biomedical literature. We apply a knowledge graph embedding method that explicitly models the uncertainty associated with literature-derived relationships and uses link prediction to generate drug repurposing hypotheses. This approach achieves high performance on a gold-standard test set of known drug indications (AUROC = 0.89) and is capable of generating novel repurposing hypotheses, which we independently validate using external literature sources and protein interaction networks. Finally, we demonstrate the ability of our model to produce explanations of its predictions.

An Approach to Multi-Objective Path Planning Optimization for Underwater Gliders.

Underwater gliders are energy-efficient vehicles that rely on changes in buoyancy in order to convert up and down movement into forward displacement. These vehicles are conceived as multi-sensor platforms, and can be used to collect ocean data for long periods in wide range areas. This endurance is achieved at the cost of low speed, which requires extensive planning to ensure vehicle safety and mission success, particularly when dealing with strong ocean currents. As gliders are often involved on missions that pursue multiple objectives (track events, reach a target point, avoid obstacles, sample specified areas, save energy), path planning requires a way to deal with several constraints at the same time; this makes glider path planning a multi-objective (MO) optimization problem. In this work, we analyse the usage of the non-dominated sorting genetic algorithm II (NSGA-II) to tackle a MO glider path planning application on a complex environment integrating 3D and time varying ocean currents. Multiple experiments using a glider kinematic simulator coupled with NSGA-II, combining different control parameters were carried out, to find the best parameter configuration that provided suitable paths for the desired mission. Ultimately, the system described in this work was able to optimize multi-objective trajectories, providing non dominated solutions. Such a planning tool could be of great interest in real mission planning, to assist glider pilots in selecting the most convenient paths for the vehicle, taking into account ocean forecasts and particular characteristics of the deployment location.

Detection of Volatile Compounds Emitted by Bacteria in Wounds Using Gas Sensors.

In this paper we analyze an experiment for the use of low-cost gas sensors intended to detect bacteria in wounds using a non-intrusive technique. Seven different genera/species of microbes tend to be present in most wound infections. Detection of these bacteria usually requires sample and laboratory testing which is costly, inconvenient and time-consuming. The validation processes for these sensors with nineteen types of microbes (1 Candida, 2 Enterococcus, 6 Staphylococcus, 1 Aeromonas, 1 Micrococcus, 2 E. coli and 6 Pseudomonas) are presented here, in which four sensors were evaluated: TGS-826 used for ammonia and amines, MQ-3 used for alcohol detection, MQ-135 for CO2 and MQ-138 for acetone detection. Validation was undertaken by studying the behavior of the sensors at different distances and gas concentrations. Preliminary results with liquid cultures of 108 CFU/mL and solid cultures of 108 CFU/cm2 of the 6 Pseudomonas aeruginosa strains revealed that the four gas sensors showed a response at a height of 5 mm. The ammonia detection response of the TGS-826 to Pseudomonas showed the highest responses for the experimental samples over the background signals, with a difference between the values ​​of up to 60 units in the solid samples and the most consistent and constant values. This could suggest that this sensor is a good detector of Pseudomonas aeruginosa, and the recording made of its values ​​could be indicative of the detection of this species. All the species revealed similar CO2 emission and a high response rate with acetone for Micrococcus, Aeromonas and Staphylococcus.

Automatized colon polyp segmentation via contour region analysis.

The increasing use of colorectal cancer screening programs has contributed to the growing number of colonoscopies performed by health centers. Hence, in recent years there has been a tendency to develop medical diagnosis support tools in order to assist specialists. This research has designed an automatized polyp detection system that allows a reduction in the rate of missed polyps that can lead to interval cancer; one of the main risks existing in colonoscopy. A characterization has therefore been made of the shape, color and curvature of edges and their regions, enabling the segmentation of polyps present in colonoscopy images. A 90.53% polyp detection rate has been achieved using the designed system, and 76.29% and 71.57% segmentation quality for the Annotated Area Covered and Dice Coefficient indicators respectively. This system aims to offer assistance with medical diagnosis that has a positive impact on patient health.

Efficient use of mobile devices for quantification of pressure injury images.

Pressure Injuries are chronic wounds that are formed due to the constriction of the soft tissues against bone prominences. In order to assess these injuries, the medical personnel carry out the evaluation and diagnosis using visual methods and manual measurements, which can be inaccurate and may generate discomfort in the patients. By using segmentation techniques, the Pressure Injuries can be extracted from an image and accurately parameterized, leading to a correct diagnosis. In general, these techniques are based on the solution of differential equations and the involved numerical methods are demanding in terms of computational resources. In previous work, we proposed a technique developed using toroidal parametric equations for image decomposition and segmentation without solving differential equations. In this paper, we present the development of a mobile application useful for the non-contact assessment of Pressure Injuries based on the toroidal decomposition from images. The usage of this technique allows us to achieve an accurate segmentation almost 8 times faster than Active Contours without Edges (ACWE) and Dynamic Contours methods. We describe the techniques and the implementation for Android devices using Python and Kivy. This application allows for the segmentation and parameterization of injuries, obtain relevant information for the diagnosis and tracking the evolution of patient's injuries.

Tissue classification and segmentation of pressure injuries using convolutional neural networks.

BACKGROUND AND OBJECTIVES: This paper presents a new approach for automatic tissue classification in pressure injuries. These wounds are localized skin damages which need frequent diagnosis and treatment. Therefore, a reliable and accurate systems for segmentation and tissue type identification are needed in order to achieve better treatment results. METHODS: Our proposed system is based on a Convolutional Neural Network (CNN) devoted to performing optimized segmentation of the different tissue types present in pressure injuries (granulation, slough, and necrotic tissues). A preprocessing step removes the flash light and creates a set of 5x5 sub-images which are used as input for the CNN network. The network output will classify every sub-image of the validation set into one of the three classes studied. RESULTS: The metrics used to evaluate our approach show an overall average classification accuracy of 92.01%, an average total weighted Dice Similarity Coefficient of 91.38%, and an average precision per class of 97.31% for granulation tissue, 96.59% for necrotic tissue, and 77.90% for slough tissue. CONCLUSIONS: Our system has been proven to make recognition of complicated structures in biomedical images feasible.

Pressure ulcer image segmentation technique through synthetic frequencies generation and contrast variation using toroidal geometry.

BACKGROUND: Pressure ulcers have become subject of study in recent years due to the treatment high costs and decreased life quality from patients. These chronic wounds are related to the global life expectancy increment, being the geriatric and physical disable patients the principal affected by this condition. Injuries diagnosis and treatment usually takes weeks or even months by medical personel. Using non-invasive techniques, such as image processing techniques, it is possible to conduct an analysis from ulcers and aid in its diagnosis. METHODS: This paper proposes a novel technique for image segmentation based on contrast changes by using synthetic frequencies obtained from the grayscale value available in each pixel of the image. These synthetic frequencies are calculated using the model of energy density over an electric field to describe a relation between a constant density and the image amplitude in a pixel. A toroidal geometry is used to decompose the image into different contrast levels by variating the synthetic frequencies. Then, the decomposed image is binarized applying Otsu's threshold allowing for obtaining the contours that describe the contrast variations. Morphological operations are used to obtain the desired segment of the image. RESULTS: The proposed technique is evaluated by synthesizing a Data Base with 51 images of pressure ulcers, provided by the Centre IGURCO. With the segmentation of these pressure ulcer images it is possible to aid in its diagnosis and treatment. To provide evidences of technique performance, digital image correlation was used as a measure, where the segments obtained using the methodology are compared with the real segments. The proposed technique is compared with two benchmarked algorithms. The results over the technique present an average correlation of 0.89 with a variation of ±0.1 and a computational time of 9.04 seconds. CONCLUSIONS: The methodology presents better segmentation results than the benchmarked algorithms using less computational time and without the need of an initial condition.