Computational Intelligence and Its Applications in Healthcare

Jitendra Kumar Verma, Sudip Paul, Prashant Johri

• Publisher: Academic Press
• Year: 2020
• Print ISBN: 9780128206041
• eBook ISBN: 9780128206195
• Topic: Biological Sciences
• Subtopic: Biotechnology
• Index: Biological Sciences

Computational Intelligence and Its Applications in Healthcare presents rapidly growing applications of computational intelligence for healthcare systems, including intelligent synthetic characters, man-machine interface, menu generators, user acceptance analysis, pictures archiving, and communication systems. Computational intelligence is the study of the design of intelligent agents, which are systems that act intelligently: they do what they think are appropriate for their circumstances and goals; they're flexible to changing environments and goals; they learn from experience; and they make appropriate choices given perceptual limitations and finite computation. Computational intelligence paradigms offer many advantages in maintaining and enhancing the field of healthcare.- Provides coverage of fuzzy logic, neural networks, evolutionary computation, learning theory, probabilistic methods, telemedicine, and robotics applications- Includes coverage of artificial intelligence and biological applications, soft computing, image and signal processing, and genetic algorithms- Presents the latest developments in computational methods in healthcare- Bridges the gap between obsolete literature and current literature

1: The impact of Internet of Things and data semantics on decision making for outpatient monitoring

Mario José Diván; María Laura Sánchez-Reynoso Data Science Research Group, Economy School, National University of La Pampa, Santa Rosa, La Pampa, Argentina

Abstract

The Internet of Things (IoT) is useful for data collection due to its portability, low cost, and the wide range of heterogeneous devices available for field deployment. Outpatients are a natural group to be monitored using IoT devices. Data semantics is a key topic in this regard, related mainly to risk prevention and support for data-driven decision making. In this chapter, an application of a measurement framework with the support of states and scenarios is introduced as a mean of homogenizing data semantics independently of the collector device. The application is focused on monitoring of outpatients engaged in physical activities outdoor. The project definition is detailed, along with the devices used in the implementation. The main contribution of this work is the integration of heterogeneous IoT devices through a measurement framework with the support of multiple scenarios and states, with data semantics used to guide the real-time data processing.

Keywords

Internet of Things; Data; Decision making; Real-time; Monitoring

Acknowledgments

This research is partially supported by the projects Res.CD 278/2016 and 312/18 of the Economics and Law School of the National University of La Pampa.

1: Introduction

The Internet of Things (IoT) has allowed the implementation of different strategies for real-time monitoring that strengthen data-driven decision making and the use of recommender systems [1]. One of the most important strengths of the IoT is the ability to integrate different hardware in a coordinated way in order to implement various data- collecting strategies. However, this strength implies heterogeneity, and data interoperability (from syntactic and semantic points of view) constitutes a current challenge in this field [2].

Health care is an ideal field for applying the IoT for patient monitoring, due to a plethora of available, cheap, usable, and wearable sensors. It is possible to coordinate all of these sensor devices under a processing strategy for orchestrating data collection, in order to prevent and avoid different risk situations [3]. Furthermore, this collecting strategy can be articulated with a recommender system, with the aim of providing suggestions to be implemented when a given situation (e.g., hypertension) is detected [4]. In this context, the recommendations imply a way to incorporate previous experience and knowledge to support decisions.

This chapter describes an integral perspective on IoT devices as a collection method in outpatient monitoring, supported by a measurement framework able to model states and scenarios, using a data- stream processing architecture for supporting data-driven decision making, along with an alternative for implementing recommendations based on the knowledge gained. Thus an application of a measurement framework with the support of states and scenarios is introduced as a mean of homogenizing the data semantics independently of the collection devices. The application focuses on outpatient monitoring of outdoor physical activities. The project definition is detailed herein, along with the devices used.

The work is organized into seven sections. Section 2 introduces some related works. Section 3 provides a synthesis of the underlying ideas associated with the measurement and evaluation (M&E) framework, along with the state and scenario modeling. Section 4 defines the M&E project, specifying the entity being monitored, the characteristics to be monitored, the associated methods, and the scenarios and states to be used in the application case. Section 5 introduces the method by which the data processing is data guided, discriminating the data meaning through tags. Section 6 discusses the application case of outpatient monitoring of outdoor physical activities. Finally, some conclusions and future work suggestions are outlined.

2: Related works

In [5] a synthesis is given of the IoT, its use in health care, and the use of machine learning for integrating data coming from a variety of sensors. The data semantics was brought onboard through a data-aware annotation strategy that allowed guiding the data meaning in relation to its processing and the application of machine-learning algorithms. Our proposal models the data differently, as a part of heterogeneous data streams flowing continuously, which are processed, analyzed, and discarded on-the-fly, with appropriate recommendations being provided.

Kaur et al. [6] proposed to use historical data jointly with data coming from IoT devices in health care in order to provide context-aware recommendations developing an active behavior. The integration proposed by the authors is interesting because it simultaneously considers IoT, historical data, the decision-making process, and the recommendation strategy. On the other hand, in our proposal, the measures coming from heterogeneous devices are completely based on a measurement and evaluation framework.

In [7] missing values and their effects on the decision-making process are addressed. The context is pregnancy monitoring, in which the vital signs need to be analyzed over 24 h every day of the week. In this environment, the authors propose a very interesting approach oriented to real-time data monitoring based on IoT devices. Our data-collecting strategy is different, as it is based on semantic annotations for identifying data meaning coming from each sensor.

In [8] an architecture is described for solving different situations related to telemedicine using IoT devices in the patient layer. The architecture contemplates the patient’s sensors as an initial approach to their current situation, while the same information is used by an algorithm for detecting the associated level of risk. Once the level of risk has been determined, the course of action and recommendations are provided immediately, avoiding the risks. Our proposal introduces metadata related to project definition for informing the data meaning related to each measure coming from a sensor. In addition, previous experience and knowledge are kept in an organizational memory organized by cases.

In [9] an integrated proposal for monitoring patients suffering obesity is introduced. A device based on the Arduino platform is used to collect data on body temperature, heart rate, blood pressure, and level of oxygen in the blood, with data jointly stored in the Arduino and on the Cloud, to provide monitoring services to the medical staff. Patients are monitored throughout the day during each activity engaged in, being under continuous monitoring. As a different approach, our proposal uses a measurement framework based on an ontology to support data meaning, jointly with its processing, storage, and use.

In [10] the possibility of image collecting is proposed, using sensor edge computing to provide a certain level of security in outpatient monitoring. In this way, outpatients contain their data as close as possible to themselves, providing a decrease in the levels of risk related to privacy and the data being viewed by others. Similarly, our proposal contemplates the use of complementary data to measurements in different formats (e.g., images, video, audio, etc.). A particular hazard of data collected in medical continuous monitoring systems is that of privacy, which continues to be a challenge to solve. Our proposal is different in that it uses various tags based on an ontology for properly discriminating each piece of data used and transmitted.

In [11] an architecture for supporting self-rehabilitation in elderly patients based on data-stream applications was introduced. The proposal discusses the monitoring of each elderly patient during rehabilitation exercises at home, while the application collects data and informs them in real time. In this context, the type of exercise to be monitored depends on the rehabilitation strategy of each outpatient, as the activity patterns of all the outpatients being monitored are not identical. Our proposal, on the other hand, specializes in measurement and evaluation projects incorporating decision criteria based on scenarios and entity states, for self-interpreting the received measurements and providing recommendations immediately when an atypical situation is detected.

3: Scenarios and states in the measurement process

The reliability of a measurement process depends on its ability to be transparently and homogeneously reproduced, with the results being comparable. For that reason, the measurement frameworks are important in defining and describing both the aim and the context of the measurement itself [12]. Our data-processing strategy specializes in measurement and evaluation projects and is based on a measurement framework [13, 14] in which each concept to be monitored is defined as an entity (e.g., an athlete) that belongs to an entity category (e.g., athletes). Each entity is contained in an environment (i.e., the context). Entities being monitored, along with their contexts, are discretely characterized through attributes and context properties, respectively. For example, the study of the athlete's activities could be addressed through the following attributes heart rate, systolic pressure, diastolic pressure, and respiration rate. The environment surrounding they could be described using the environmental temperature and humidity as context properties.

The number of attributes defines the dimensionality of the entity, while the number of context properties does the same for the environment. Thus, entity and context are described through the measurements related only to the indicated attributes and not others.

Each attribute and context property has its own behavior characterized by means of the data series, jointly varying throughout the measurement process. Because the variation range of attributes and context properties can be known (e.g., heart rate in an athlete depending on the age) or eventually estimated, different variation intervals can be established for them. Of course, each interval definition (be it for attributes or context properties) implies the assistance of an exper...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• Contributors
• 1: The impact of Internet of Things and data semantics on decision making for outpatient monitoring
• 2: Deep-learning approaches for health care: Patients in intensive care
• 3: Brain MRI image segmentation using nature-inspired Black Hole metaheuristic clustering approach
• 4: Blockchain for public health: Technology, applications, and a case study
• 5: Compression and multiplexing of medical images using optical image processing
• 6: Analysis of skin lesions using machine learning techniques
• 7: Computational intelligence using ontology—A case study on the knowledge representation in a clinical decision support system
• 8: Neural network-based abnormality detection for electrocardiogram time signals
• 9: Machine learning approaches for acetic acid test based uterine cervix image analysis
• 10: Convolutional neural network for biomedical applications
• 11: Alzheimer’s disease classification using deep learning
• 12: Diabetic retinopathy identification using autoML
• 13: Knowledge-based systems in medical applications
• 14: Convolution neural network-based feature learning model for EEG-based driver alert/drowsy state detection
• 15: Analysis on the prediction of central line-associated bloodstream infections (CLABSI) using deep neural network classification
• Index