Ebook: Roles and Challenges of Semantic Intelligence in Healthcare Cognitive Computing

This book delves into emerging fields of science and technology, where cognitive computing techniques offer effective solutions that are poised to have a significant impact in the foreseeable future. The fusion of semantic intelligence, machine learning, big data, artificial intelligence, analytics, and natural language processing necessitates collaboration among diverse disciplines, encompassing science, technology, business, and the humanities. This interdisciplinary approach is crucial for the collection and comprehension of information, using various senses and learning from experience. The proliferation of these technologies also presents numerous opportunities to enhance efficiency in healthcare. In this transformative era, errors will be minimised, data insights quicker and more sophisticated, and the system will be capable of proposing personalised care models.

The objective of this book is to assess the current landscape, and identify the roles and challenges encountered in the integration of cognitive computing techniques into the widespread adoption of innovative smart healthcare solutions. The healthcare domain is characterised by an array of data types and formats, including drug dosages, molecular structures, text, numerical data, and images, and healthcare systems are continuously acquiring novel data from cutting-edge technologies such as wearable devices. Semantic intelligence technologies are taking centre stage in the development of intelligent, knowledge-based systems, as they enable machines to intelligently integrate and process resources within relevant contexts. This includes a range of semantic intelligence technologies such as artificial intelligence, machine learning, deep learning, the internet of things, and the hybrid methodologies combining these approaches.

The book’s content spans a wide spectrum of cutting-edge topics and emerging trends, reflecting the multidisciplinary nature of healthcare and cognitive computing, from digital twins to eXplainable AI, from AI-based decision support systems in intensive care to semantics for culinary health care and the semantic internet of things, and from natural language processing to deep learning and graph models.

Each chapter is the result of collaboration by experts from various domains and provides a detailed overview of the potential that new technologies in the field of healthcare currently offer or will offer in the near future. We believe that the book will inspire new ideas and facilitate collaboration among the different disciplines involved. The contributions of the individual chapters are as follows.

The first chapter, “Digital twins, digital triplets, and eXplainable AI, in precision health”, explores precision health, focusing on the principles of preventing and treating diseases using digital twins and digital triplets. Digital twins are digital representations of physical objects which facilitate machine learning and knowledge mining, enabling in-silico simulations of health and molecular states and aiding evidence-based medicine. In contrast, digital triplets provide semantic intelligence by clustering similar objects to uncover hidden knowledge and interrelationships, facilitating accurate clinical and medical predictions. The chapter categorises digital twins into person phenotype digital twins, person genotype digital twins, and physicians’ brain digital twins, each of which serve a distinct purposes in advancing precision health.

The second chapter, “AI-based decision support systems in intensive care” delves into the pivotal role of intensive care units (ICUs) in modern healthcare, highlighting the challenges they face, which include high workload, complex decision-making, and managing vast amounts of healthcare data while also considering the individual needs of patients. It explores how AI-based decision support systems (DSS) are advancing healthcare by analysing patient data to make personalised care recommendations. Various AI models, such as regression, decision trees, and neural networks, are discussed. However, the chapter underscores the need for careful evaluation, ethical considerations, and clarification as regards responsibility and accountability for AI-DSS in clinical practice. While AI-DSS has potential, it also faces hurdles like data quality and scepticism among healthcare personnel. Standardisation and regulatory frameworks are deemed crucial for widespread AI adoption in ICUs.

The third chapter, “Semantics for culinary health care” explores the underutilised potential of semantic-web technologies in computational creativity, specifically in culinary healthcare. It discusses the idea of automating recipe generation by extracting and annotating semantic information from existing recipes, representing them as knowledge graphlets. The challenge lies in replicating the complexity of human culinary expertise. The chapter highlights the gap in leveraging the semantic web for creative recipe generation and emphasises the importance of structured approaches. By using semantic information, such as ontologies and knowledge graphs, the chapter envisions the creation of diverse, health-conscious recipes with innovative combinations. It suggests that combining the semantic web with AI can revolutionise the culinary landscape, making it more accessible, innovative, and health-conscious. The chapter is organised to explore the relationship between food and health, automated recipe generation, the role of the semantic web, and concludes with key findings.

The fourth chapter, entitled “Exploring the determinants of the semantic internet of things in healthcare” introduces the concept of the semantic internet of things (SIoT), an extension of the internet of things (IoT), which leverages semantic technologies to enhance interoperability and data understanding among connected devices. The SIoT is particularly valuable in the context of smart health, where it establishes a common language and framework for sharing and integrating health-related data from diverse sources. By doing so, it can enhance the accuracy and efficiency of health monitoring and decision making, enabling more personalised and effective healthcare interventions. The chapter provides an overview of SIoT’s core concepts, outlines its focus areas, and discusses its various requirements, challenges, and applications. It includes a literature review of healthcare systems employing SIoT, highlighting cases of its successful use and challenges. Additionally, the chapter explores the utilisation of semantic technologies in healthcare IoT, discussing new projects, frameworks, and specific ontologies designed to address issues related to semantic interoperability and data heterogeneity. This comprehensive overview serves as a valuable resource and guideline for future research in this field.

The fifth chapter, “Natural language processing for ontology development in IOT-enabled smart healthcare” delves into the challenges of managing vast volumes of unstructured narrative data in the healthcare industry. It highlights how ontologies, which provide formal specifications for health data, enable machine-readable representations and facilitate the sharing and reuse of health information. With the growth of interconnected systems and the internet of things (IoT), healthcare services have evolved, and semantic web technologies and ontologies play a crucial role in ensuring interoperability within IoT ecosystems. The chapter emphasises the importance of natural language processing (NLP) in ontology development for healthcare, as it helps convert unstructured health data into meaningful representations. NLP techniques are essential for cognitive computing systems, enhancing the quality of healthcare services while reducing costs. The chapter explores NLP methods in the context of IoT-enabled healthcare, presenting principles, methodologies for ontology creation using NLP, and the relationship between NLP and the semantic web. It also examines recent studies, applications, opportunities, and challenges in this field.

The sixth chapter is entitled “On the modelling of academic profiles: a semantic-web case of study”, and focuses on enhancing web search results by incorporating semantic-web elements such as ontologies and annotations. It introduces an ontology for academic profiles, and outlines the development of a software system for querying academic profile information. The system comprises a mobile application for Android and a web service for knowledge base access. A case study involving academic profiles at the Faculty of Engineering and Sciences of the Autonomous University of Tamaulipas is also presented. This work aims to improve information retrieval by utilising semantic technologies in academia.

The seventh chapter, “Deep learning to explore the correlation between screen time and depression through analysis of the NHANES dataset” explores the impact, particularly among teenagers, of increased screen time due to factors like online education and entertainment. Excessive screen time, including the use of smart phones, laptops, and televisions, can lead to depressive symptoms, cognitive behaviour issues, loneliness, personality degradation, and even suicidal thoughts in teenagers. The chapter investigates the feasibility of detecting depression based on screen usage and highlights the importance of promoting healthier digital habits. The study uses the NHANES dataset, which includes standardised questionnaires like the Hamilton Rating Scale for Depression (HAM-D), to assess depressive symptoms. The chapter focuses on the HAM-D scale and employs a generalised regression neural network (GRNN) depression-detection model based on screen time. This model achieves an impressive accuracy rate of 98.55%.

The eighth chapter, “Application of factor graph model in stress detection using social network messages” explores the use of social network data to proactively detect psychological stress. With the increasing use of social platforms, users regularly share their daily activities and engage with friends online. The research collects real-time data to investigate the correlation between users’ stress levels and their social interactions. It employs a comprehensive set of stress-related attributes, including textual, visual, and social aspects, to gauge stress levels in users. A novel model is proposed that considers the stress states of both users and their friends, improving detection performance. An accessible website is created to allow users to assess their stress levels and access stress management resources. The model’s effectiveness in detecting stress is demonstrated through experiments, and proactive care and support is offered for individuals facing psychological stress.

Precision health is about preventing, predicting, and treating diseases precisely with the principles of the right care at the right time for the right patient. Precision health is expected to help increase health equity in general. Digital Twins and Digital Triplets can significantly help in meeting the goals of precision health. In this chapter we present digital twins and digital triplets and their roles in realizing precision health. Digital twin is the digital representation of a physical object in the digital space. In the Precision Health context, digital twins are indeed enablers of machine learning and knowledge mining. It is also useful for the in-silico simulation of a person’s phenotype (health states) and genotype (molecular states) to realize evidence-based medicine. Moreover, using probabilistic graph model and neuro-symbolic AI, digital twins will be useful in mitigating physician’s knowledge gaps or decision gaps to achieve value-based care. In contrast to Digital twins, Digital triplet is the semantic intelligence about the object. Digital triplets capture the semantics by placing semantically similar objects close together in the vector embedding space. This semantic intelligence helps cognition and discover hidden and unknown knowledge and their interrelationships to make accurate clinical and medical predictions. We group digital twins in three major categories, namely, Person Phenotype Digital Twin, Person Genotype Digital Twin, and Physicians’ Brain Digital Twin. Person phenotype digital twin relates to all observable properties of a person and a population. Person genotype digital twin helps understand the molecular properties of a person and a population. Physicians brain digital twin is the doctors’ brain with actionable biomedical knowledge in the virtual space.

Intensive Care Units (ICUs) serve a critical role in providing specialized care to critically ill patients in modern healthcare. The ICU environment poses numerous challenges, including high workload, complex decision-making, management of large healthcare data, and the consideration of individual patient needs and preferences. Decision support systems (DSS) in healthcare aim to improve care and aid decision-making by providing person-specific information and recommendations. Recent advancements in Artificial Intelligence (AI) have enabled the development of AI-based DSS, which analyze extensive medical data to identify patterns, make predictions, and personalize care. Various AI models and approaches, such as regression models, decision trees, random forest models, support vector machines, and neural networks have been employed to analyze patient data and help with decision making in the ICU. Evaluation of AI tool performance employs metrics such as accuracy, precision, sensitivity, and specificity, with validation against external databases being necessary. Gaining insights into decision-making factors, integrating subjective factors, and addressing ethical concerns are essential for the acceptance and deployment of AI-DSS in clinical practice. Clarification regarding responsibility, accountability, and the ability for clinicians to override AI-DSS recommendations is necessary to ensure appropriate utilization. While efforts have been made to evaluate AI-DSS in the ICU, most applications remain retrospective and lack clear evidence of improved clinician performance or patient outcomes. The majority of AI models rely on supervised learning for detection and identification tasks, but their efficacy is limited by the scarcity of high-quality data and inadequate consideration of human factors.. Furhtermore, the implementation of AI in the ICU still faces limitations and skepticism from healthcare personnel. Transparent explanations of modeling methods and operational procedures are necessary to instill clinician confidence. Addressing challenges such as the lack of standardized ICU databases and regulatory frameworks is pivotal for widespread adoption of AI in the ICU.

In this chapter, we demonstrate the potential of semantic web and knowledge graphs in the context of culinary healthcare. We explore the idea of annotating text-based recipes with cooking semantics, rendering the underlying concepts machine-readable through the application of AI and NLP techniques. This annotated data can be harnessed to generate innovative, nutritionally sound recipes, with optimization criteria customized to prioritize health-related considerations. The concept of Semantic Recipe Generation offers promising applications, such as catering to individuals with specific dietary requirements and promoting ecofriendly cooking practices. We envision the possibility of extending this approach to develop a robust system that contributes to improving the well-being.

The Semantic Internet of Things (SIoT) is an Internet of Things (IoT) extension that uses semantic technologies to improve interoperability and understanding of data generated by connected devices. The SIoT can help in smart health by providing a common language and framework for exchanging and integrating health-related data from various sources. This has the potential to improve the accuracy and efficiency of health monitoring and decision-making, allowing for more personalized and effective health interventions. This chapter outlines the Semantic Internet of Things’ core concepts, specifies its focus areas, and lists its various requirements, difficulties, and uses. Through the presentation of a literature review on Semantic Internet of Things healthcare systems, including challenges, prospects, and current work, we highlight the best outcomes for semantics among IoT systems in healthcare, which can serve as a guideline for future research. The chapter also provides an overview of the use of semantic technologies in healthcare IoT, including new projects, frameworks, and specific ontologies that use semantic technologies to address semantic interoperability and heterogeneity issues.

The healthcare industry is a knowledge-driven system and continuously expanding with enormous volumes of narrative data that are typically stored in unstructured and non-standardized formats. Therefore, it is challenging for systems to manage massive amounts of narrative data, comprehend the contents of these data, find relevant and useful healthcare information, and make decisions. In the healthcare domain, ontologies provide a formal specification of health data for knowledge representation and data integration. Therefore, ontologies enable the representation of health information in a machine-readable form and allow this information to be shared, reused, and used to make deductions. In recent years, modern healthcare services have been significantly affected by the growth of the interconnectedness of systems and the improvements in the Internet of Things (IoT). Moreover, wearables and implantable appliances enhance people’s quality of life and ease their life by performing continuous health monitoring. IoT technologies facilitate data flow across multiple entities in healthcare systems and deal with different data formats. Semantic Web and ontologies provide interoperability among IoT ecosystems by describing concepts and relationships between different entities. Nevertheless, extracting concepts and relationships from the healthcare domain is one of the most important processes in ontology development. Therefore, Semantic Web technologies benefit from Natural Language Processing (NLP) technologies to convert unstructured health data to meaningful representations. Thus, the structured and unstructured data can be combined by integrating NLP and Semantic Web technologies. Hence, NLP is a fundamental capability for cognitive computing systems and is frequently characterized as a behavioral technology that assists computers in understanding and comprehending human language. In the health domain, NLP techniques are used to gather unstructured healthcare data, examine its grammatical structure, and ascertain its meaning. Consequently, NLP techniques increase the quality of healthcare services while reducing costs. This chapter presents the NLP methods applied to the IoT-enabled healthcare domain in the scope of semantic intelligence. For this purpose, the main principles of NLP for the Semantic Web, and methodologies for creating ontologies by utilizing the NLP techniques to process heterogeneous data sources are presented. Also, the relationship between NLP and Semantic Web is explored within the context of ontology creation and population for IoT-based healthcare systems. Finally, the recent studies, applications, opportunities, and challenges in the related field are examined.

Searching for information on the Web is a crucial activity in which, to obtain relevant results, elements that describe the context of the information must be added. This can be achieved by integrating Semantic Web elements such as ontologies (vocabulary related to a topic) and annotations (descriptions of resources using an ontology) into software systems. This chapter proposes the design and implementation of an ontology in the domain of academic profiles. It also describes the design and implementation of a software system for querying information on academic profiles using the ontology as a knowledge base. The software system consists of two main components: 1) mobile application for the Android platform and 2) web service for access and management of the knowledge base. The academic profiles of the Faculty of Engineering and Sciences of the Autonomous University of Tamaulipas are described as a case study.

The explosion in the number of digital gadgets and the increasing integration of technology into daily life have led to a significant rise in screen time usage across various age groups. Also, the pandemic has moved us indoors leading every work to be done online. The screen time including the smart phones, Laptops and Televisions have increased among the teenagers with the advent of online education becoming a necessity and exceed the American Academy of Pediatrics recommendations. The time devoted to social media, online gaming, and entertainment platforms leads to depressive symptoms. The blue light emanating from these devices can cause intensive damage to the metal health of teenagers. The hazards of using these devices for a longer time has an influence both on their behaviour and their rational thinking. The effect on cognitive behaviour is reflected in their mental health like loneliness, degradation of personality level and suicidal thoughts. There arises a debate to find if usage of digital gadgets for a long time has depressive symptoms in teenagers. In this chapter the feasibility of detecting depression based on the screen time usage is explored and an awareness is created to promote healthier digital habits. The NHANES, dataset contains various standardized questionnaires designed to evaluate the severity of depressive symptoms is considered for analysis. Some widely used depression scales include Beck Depression Inventory (BDI), Hamilton Rating Scale for Depression (HAM-D) etc., In this chapter, we have chosen HAM-D to assess the severity of depression in an individual. Higher scores indicate that the person is in severe depression and needs treatment. A Generalized regression neural network (GRNN) depression detection model based on screen time usage is proposed and the model is compared with Multilayer Neural Network (MLNN) Model. The proposed model gives an accuracy of 98.55

Psychological stress poses a significant threat to people’s well-being, and it can be challenging to identify and address stress in a timely manner. With an increase in the usage of social network platforms, people have become accustomed to share their daily routine activities and engage with their friends online. This presents an opportunity to leverage the data from these online social networks for detecting stress proactively. In this research work, real time dataset is collected to investigate the correlation between users’ stress level and their social interactions. A comprehensive set of stress-related attributes, encompassing textual, visual, and social aspects serve as indicators of stress levels in users. A novel model is proposed in this study that can utilize these stress-related attributes to improve detection performance. The model considers the stress states of both the users and their friends in the social network. By analyzing user interactions and shared content, stress levels can be accurately gauged. To make findings of this study accessible to the public, a website is created allowing users to assess their stress levels and explore related activities. By using this platform, individuals can gain insights into their own stress rates and access resources to manage stress effectively. Through the experiments and analysis, the effectiveness of the proposed model in detecting stress is demonstrated. The integration of social network data and stress-related attributes significantly enhances the accuracy of stress detection, enabling proactive care and support for individuals facing psychological stress.