From vision to text: A comprehensive review of natural image captioning in medical diagnosis and radiology report generation.

Natural Image Captioning (NIC) is an interdisciplinary research area that lies within the intersection of Computer Vision (CV) and Natural Language Processing (NLP). Several works have been presented on the subject, ranging from the early template-based approaches to the more recent deep learning-based methods. This paper conducts a survey in the area of NIC, especially focusing on its applications for Medical Image Captioning (MIC) and Diagnostic Captioning (DC) in the field of radiology. A review of the state-of-the-art is conducted summarizing key research works in NIC and DC to provide a wide overview on the subject. These works include existing NIC and MIC models, datasets, evaluation metrics, and previous reviews in the specialized literature. The revised work is thoroughly analyzed and discussed, highlighting the limitations of existing approaches and their potential implications in real clinical practice. Similarly, future potential research lines are outlined on the basis of the detected limitations.

Unlocking the power of AI models: exploring protein folding prediction through comparative analysis.

Protein structure determination has made progress with the aid of deep learning models, enabling the prediction of protein folding from protein sequences. However, obtaining accurate predictions becomes essential in certain cases where the protein structure remains undescribed. This is particularly challenging when dealing with rare, diverse structures and complex sample preparation. Different metrics assess prediction reliability and offer insights into result strength, providing a comprehensive understanding of protein structure by combining different models. In a previous study, two proteins named ARM58 and ARM56 were investigated. These proteins contain four domains of unknown function and are present in Leishmania spp. ARM refers to an antimony resistance marker. The study's main objective is to assess the accuracy of the model's predictions, thereby providing insights into the complexities and supporting metrics underlying these findings. The analysis also extends to the comparison of predictions obtained from other species and organisms. Notably, one of these proteins shares an ortholog with Trypanosoma cruzi and Trypanosoma brucei, leading further significance to our analysis. This attempt underscored the importance of evaluating the diverse outputs from deep learning models, facilitating comparisons across different organisms and proteins. This becomes particularly pertinent in cases where no previous structural information is available.

Uncovering hidden therapeutic indications through drug repurposing with graph neural networks and heterogeneous data.

Drug repurposing has gained the attention of many in the recent years. The practice of repurposing existing drugs for new therapeutic uses helps to simplify the drug discovery process, which in turn reduces the costs and risks that are associated with de novo development. Representing biomedical data in the form of a graph is a simple and effective method to depict the underlying structure of the information. Using deep neural networks in combination with this data represents a promising approach to address drug repurposing. This paper presents BEHOR a more comprehensive version of the REDIRECTION model, which was previously presented. Both versions utilize the DISNET biomedical graph as the primary source of information, providing the model with extensive and intricate data to tackle the drug repurposing challenge. This new version's results for the reported metrics in the RepoDB test are 0.9604 for AUROC and 0.9518 for AUPRC. Additionally, a discussion is provided regarding some of the novel predictions to demonstrate the reliability of the model. The authors believe that BEHOR holds promise for generating drug repurposing hypotheses and could greatly benefit the field.

Contribution of Dynamic and Genetic Tests for Short Stature Diagnosing: A Case Report.

Genetic tests have led to the discovery of many novel genetic variants related to growth failure, but the clinical significance of some results is not always easy to establish. The aim of this report is to describe both clinical phenotype and genetic characteristics in an adult patient with short stature associated with a homozygous variant in disintegrin and metalloproteinase with thrombospondin motifs type 17 gene (ADAMTS17) combined with a homozygous variant in the GH secretagogue receptor (GHS-R). The index case had severe short stature (SS) (-3.0 SD), small hands and feet, associated with eye disturbances. Genetic tests revealed homozygous compounds for ADAMTS17 responsible for Weill-Marchesani-like syndrome but a homozygous variant in GHS-R was also detected. Dynamic stimulation with an insulin tolerance test showed a normal elevation of GH, while the GH response to macimorelin stimulus was totally flattened. We show the implication of the GHS-R variant and review the molecular mechanisms of both entities. These results allowed us to better interpret the phenotypic spectrum, associated co-morbidities, its implications in dynamic tests, genetic counselling and treatment options not only to the index case but also for her relatives.

Discovering Hidden Mental States in Open Multi-Agent Systems by Leveraging Multi-Protocol Regularities with Machine Learning.

The agent paradigm and multi-agent systems are a perfect match for the design of smart cities because of some of their essential features such as decentralization, openness, and heterogeneity. However, these major advantages also come at a great cost. Since agents' mental states are hidden when the implementation is not known and available, intelligent services of smart cities cannot leverage information from them. We contribute with a proposal for the analysis and prediction of hidden agents' mental states in a multi-agent system using machine learning methods that learn from past agents' interactions. The approach employs agent communication languages, which is a core property of these multi-agent systems, to infer theories and models about agents' mental states that are not accessible in an open system. These mental state models can be used on their own or combined to build protocol models, allowing agents (and their developers) to predict future agents' behavior for various tasks such as testing and debugging them or making communications more efficient, which is essential in an ambient intelligence environment. This paper's main contribution is to explore the problem of building these agents' mental state models not from one, but from several interaction protocols, even when the protocols could have different purposes and provide distinct ambient intelligence services.

Prediction and Decision-Making in Intelligent Environments Supported by Knowledge Graphs, A Systematic Review.

Ambient Intelligence is currently a lively application domain of Artificial Intelligence and has become the central subject of multiple initiatives worldwide. Several approaches inside this domain make use of knowledge bases or knowledge graphs, both previously existing and ad hoc. This form of representation allows heterogeneous data gathered from diverse sources to be contextualized and combined to create relevant information for intelligent systems, usually following higher level constraints defined by an ontology. In this work, we conduct a systematic review of the existing usages of knowledge bases in intelligent environments, as well as an in-depth study of the predictive and decision-making models employed. Finally, we present a use case for smart homes and illustrate the use and advantages of Knowledge Graph Embeddings in this context.

A Participatory Agent-Based Simulation for Indoor Evacuation Supported by Google Glass.

Indoor evacuation systems are needed for rescue and safety management. One of the challenges is to provide users with personalized evacuation routes in real time. To this end, this project aims at exploring the possibilities of Google Glass technology for participatory multiagent indoor evacuation simulations. Participatory multiagent simulation combines scenario-guided agents and humans equipped with Google Glass that coexist in a shared virtual space and jointly perform simulations. The paper proposes an architecture for participatory multiagent simulation in order to combine devices (Google Glass and/or smartphones) with an agent-based social simulator and indoor tracking services.

Towards a holistic framework for the evaluation of emergency plans in indoor environments.

One of the most promising fields for ambient intelligence is the implementation of intelligent emergency plans. Because the use of drills and living labs cannot reproduce social behaviors, such as panic attacks, that strongly affect these plans, the use of agent-based social simulation provides an approach to evaluate these plans more thoroughly. (1) The hypothesis presented in this paper is that there has been little interest in describing the key modules that these simulators must include, such as formally represented knowledge and a realistic simulated sensor model, and especially in providing researchers with tools to reuse, extend and interconnect modules from different works. This lack of interest hinders researchers from achieving a holistic framework for evaluating emergency plans and forces them to reconsider and to implement the same components from scratch over and over. In addition to supporting this hypothesis by considering over 150 simulators, this paper: (2) defines the main modules identified and proposes the use of semantic web technologies as a cornerstone for the aforementioned holistic framework; (3) provides a basic methodology to achieve the framework; (4) identifies the main challenges; and (5) presents an open and free software tool to hint at the potential of such a holistic view of emergency plan evaluation in indoor environments.

An approach for representing sensor data to validate alerts in Ambient Assisted Living.

The mainstream of research in Ambient Assisted Living (AAL) is devoted to developing intelligent systems for processing the data collected through artificial sensing. Besides, there are other elements that must be considered to foster the adoption of AAL solutions in real environments. In this paper we focus on the problem of designing interfaces among caregivers and AAL systems. We present an alert management tool that supports carers in their task of validating alarms raised by the system. It generates text-based explanations--obtained through an argumentation process--of the causes leading to alarm activation along with graphical sensor information and 3D models, thus offering complementary types of information. Moreover, a guideline to use the tool when validating alerts is also provided. Finally, the functionality of the proposed tool is demonstrated through two real cases of alert.