Artificial Intelligence Applications for Health Care

Mitul Kumar Ahirwal, Narendra D. Londhe, Anil Kumar

• Verlag: CRC Press
• Jahr: 2022
• ISBN:
  9781000570342
• Auflage: 1
• Thema: Business
• Thema: Industria dei servizi

This book takes an interdisciplinary approach by covering topics on health care and artificial intelligence. Data sets related to biomedical signals (ECG, EEG, EMG) and images (X-rays, MRI, CT) are explored, analyzed, and processed through different computation intelligence methods. Applications of computational intelligence techniques like artificial and deep neural networks, swarm optimization, expert systems, decision support systems, clustering, and classification techniques on medial datasets are explained. Survey of medical signals, medial images, and computation intelligence methods are also provided in this book.

Key Features

• Covers computational Intelligence techniques like artificial neural networks, deep neural networks, and optimization algorithms for Healthcare systems
• Provides easy understanding for concepts like signal and image filtering techniques
• Includes discussion over data preprocessing and classification problems
• Details studies with medical signal (ECG, EEG, EMG) and image (X-ray, FMRI, CT) datasets
• Describes evolution parameters such as accuracy, precision, and recall etc.

This book is aimed at researchers and graduate students in medical signal and image processing, machine and deep learning, and healthcare technologies.

1 A Survey of Machine Learning in Healthcare

S. Sathyanarayanan

Sri Sathya Sai University for Human Excellence, Gulbarga, India

Sanjay Chitnis

Dayananda Sagar University, Bangalore, India

DOI: 10.1201/9781003241409-1

CONTENTS

1. 1.1 Introduction
2. 1.2 Artificial Intelligence
3. 1.2.1 Machine Learning
4. 1.2.1.1 Steps in Developing an ML System
5. 1.2.1.2 Types of Machine Learning
6. 1.2.2 Deep Learning
7. 1.2.3 The Major Types of DL
8. 1.3 Applications of ML in Healthcare
9. 1.3.1 Cardiovascular Diseases
10. 1.3.2 Medical Imaging
11. 1.3.3 Drug Discovery/Manufacturing
12. 1.3.4 Electronic Health Records
13. 1.3.5 Clinical Decision Support System
14. 1.3.6 Surgical Robotics
15. 1.3.7 Precision Medicine
16. 1.3.8 Population Health Management
17. 1.3.9 mHealth and Smart Devices
18. 1.3.10 AI for Tackling Pandemic
19. 1.4 ML Use Cases in Healthcare
20. 1.5 Limitations and Challenges in Adoption of AI in Healthcare
21. 1.6 Conclusion
22. Acknowledgements
23. References

1.1 Introduction

The advanced medical equipment used in secondary and tertiary healthcare centres is considerably expensive, and obtaining quality healthcare has become difficult and unaffordable for majority of the population. Hence, the application of Machine Learning (ML) in healthcare domains to automate various procedures is expected to help accelerate diagnostic procedures, reduce cost and increase access to quality care for general population. ML can be used in the analysis of healthcare data for faster diagnosis. ML-enabled systems can also be used for prediction by analysing the vast amount of healthcare data to enable preventive measures in time. ML in biomedical technologies will help in the development of smaller, portable, easy-to-use, cloud-enabled and affordable devices, which can be used by the medical staff at primary healthcare centres; thereby, reducing the workload of doctors and eliminating the need for advanced diagnostic equipment for most of the patients. This will also make expensive resources more easily available for people in need. Thus, ML in healthcare could provide greater access to quality care with minimal hindrances.

This chapter reviews some of the ongoing research in this area after a brief tutorial of ML and also briefly describes few ML-enabled devices already in use.

1.2 Artificial Intelligence

Traditionally, computers had to be programmed by humans. This limited its application in areas with set rules and specific protocols. John McCarthy, an American computer scientist, defined AI as the science and engineering of making intelligent machines [1]. AI is used to describe systems that mimic “cognitive” functions that are associated with humans, such as problem-solving skills and learning [2]. Narrow AI systems are those that carry out specific tasks, for example the AI systems currently being used in healthcare domain. These systems are common, for example virtual assistants on smart phones and recommendation systems. Artificial General Intelligence (AGI) systems are capable of learning different tasks like humans. However, such systems do not exist as of now. The different types of AI systems based on the type of work they accomplish are briefly discussed here and depicted in Figure 1.1.

Figure 1.1 Types of AI.

Knowledge Representation and Reasoning (KR & R or KR): This is a type of AI wherein knowledge representation deals with representing information in the form of symbols and propositions. It makes use of reasoning as the method of deducing facts from the data, for example, computer-aided diagnosis.

Automated Planning and Scheduling: This is also called AI planning. It considers the time and resource constraints and generates a series of actions in sequence to achieve the objective.

Machine Learning: This gives computers the capability to learn from data without being programmed manually.

Natural Language Processing (NLP): This aims to build systems that can make sense out of the raw natural language data as input. An example is Siri in Apple devices.

Computer Vision: This enables computers to process image and video data and see, observe and make sense out of it. For example, computer vision helps in detecting patterns in various medical imaging data to help in diagnosing diseases.

Robotics: A robot is an autonomous system or a system with external control with the capability to perceive its environment, make decisions and perform actions in the real world. Robotics is a field that makes use of various branches of science and engineering to build autonomous robots.

Artificial General Intelligence: The goal of AI as a field is to move towards Artificial General Intelligence (AGI). A system having AGI implies that the system can match humans in intellectual capacity. The examples considered so far are examples of narrow AI.

1.2.1 Machine Learning

ML is a subfield of computer science that provides computers the ability to learn to perform a task on its own from experience without being explicitly programmed. Tom Mitchell defined learning precisely with the following definition [3]: “A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E”.

Such systems learn from the data provided and give a prediction for new data. The conventional ML algorithms are dependent on the availability of data for effectiveness. For example, a price prediction algorithm for houses may give a prediction for the house if the location, size, etc. are given. Each information included in the representation of the house for sale is known as a feature. The ML system learns how each of these features of the house correlates with various outcomes. The conventional programming systems give output based on the program and input data. The ML system gives a model as its output based on the input data and the various outcomes which is fed to it. This is illustrated in Figure 1.2.

Figure 1.2 Difference between traditional programming system and machine learning system.

1.2.1.1 Steps in Developing an ML System

A project to develop an ML-enabled system involves multiple phases before it can be put to use.

Data Collection and Labelling: The first step in ML is the collection of data, including the detailed description of the data involved that will help in understanding and identifying the type of data required for the ML model. The process of detecting and tagging data samples is called data labelling.

Pre-processing: After data is collected, it has to be pre-processed, which may involve deduplication, normalisation and error correction to make the data suitable for use.

Feature Engineering: A feature is an attribute or a measurable property shared by all the independent units of the data on which analysis or prediction is to be done [4].

For example, in property prices data, the area of the house, the number of rooms and the age of the property can be considered as features. The collection of data of all the houses sold may be considered as...