Top

2021 | Book

Read chapter Read first chapter

Advances in Artificial Intelligence, Computation, and Data Science

For Medicine and Life Science

Editors: Tuan D. Pham, Hong Yan, Muhammad W. Ashraf, Folke Sjöberg

Publisher: Springer International Publishing

Book Series : Computational Biology

Part of: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Table of Contents

About this book

Artificial intelligence (AI) has become pervasive in most areas of research and applications. While computation can significantly reduce mental efforts for complex problem solving, effective computer algorithms allow continuous improvement of AI tools to handle complexity—in both time and memory requirements—for machine learning in large datasets. Meanwhile, data science is an evolving scientific discipline that strives to overcome the hindrance of traditional skills that are too limited to enable scientific discovery when leveraging research outcomes. Solutions to many problems in medicine and life science, which cannot be answered by these conventional approaches, are urgently needed for society.

This edited book attempts to report recent advances in the complementary domains of AI, computation, and data science with applications in medicine and life science. The benefits to the reader are manifold as researchers from similar or different fields can be aware of advanced developments and novel applications that can be useful for either immediate implementations or future scientific pursuit.

Features:

Considers recent advances in AI, computation, and data science for solving complex problems in medicine, physiology, biology, chemistry, and biochemistryProvides recent developments in three evolving key areas and their complementary combinations: AI, computation, and data scienceReports on applications in medicine and physiology, including cancer, neuroscience, and digital pathologyExamines applications in life science, including systems biology, biochemistry, and even food technology

This unique book, representing research from a team of international contributors, has not only real utility in academia for those in the medical and life sciences communities, but also a much wider readership from industry, science, and other areas of technology and education.

Frontmatter

Bioinformatics

Frontmatter

1. Intelligent Learning and Verification of Biological Networks

Abstract

Machine learning and model checking are two types of intelligent computing techniques that have been widely used to study different complicated systems nowadays. It is well-known that the cellular functions and biological processes are strictly regulated by different biological networks, for example, signaling pathways and gene regulatory networks. The pathogenesis of cancers is associated with the dysfunctions of some regulatory networks or signaling pathways. A comprehensive understanding of the biological networks could identify cellular signatures and uncover hidden pathological mechanisms, and help develop targeted therapies for cancers and other diseases. In order to correctly reconstruct biological networks, statisticians and computer scientists have been motivated to develop many intelligent methods, but it is still a challenging task due to the complexity of the biological system and the curse of dimensionality of the high-dimensional biological data. In this work, we will review different machine learning algorithms and formal verification (model checking) techniques that have been proposed and applied in our previous work and discuss how to integrate these computational methods together to intelligently infer and verify complex biological networks from biological data. The advantages and disadvantages of these methods are also discussed in this work.

Helen Richards, Yunge Wang, Tong Si, Hao Zhang, Haijun Gong

Chapter 2. Differential Expression Analysis of RNA-Seq Data and Co-expression Networks

Abstract

At present, RNA-seq has become the most common and powerful platform in the study of transcriptomes. A major goal of RNA-seq analysis is the identification of genes and molecular pathways which are differentially expressed in two altered situations. Such difference in expression profiles might be linked with changes in biology giving an indication for further intense investigation. Generally, the traditional statistical methods used in the study of differential expression analysis of gene profiles are restricted to individual genes and do not provide any information regarding interactivities of genes contributing to a certain biological system. This need led the scientists to develop new computational methods to identify such interactions of genes. The most common approach used to study gene-set interactivities is gene network inference. Co-expression gene networks are the correlation-based networks which are commonly used to identify the set of genes significantly involved in the occurrence or presence of a particular biological process. This chapter describes a basic procedure of an RNA-seq analysis along with a brief description about the techniques used in the analysis: an illustration on a real data set is also shown. In addition, a basic pipeline is presented to elucidate how to construct a co-expression network and detect modules from the RNA-seq data.

Sana Javed

3. Learning Biomedical Networks: Toward Data-Informed Clinical Decision and Therapy

Abstract

Precision medicine has emerged to tailor clinical decisions based on patient genetic features in a personalized healthcare perspective. The ultimate goal is to drive disease diagnosis and treatment selection based on the patient molecular profiles, usually given by large volumes of data, which is intrinsically high-dimensional, heterogeneous, noisy, and incomplete. Along with the notable improvement of experimental technologies, statistical learning has accompanied the associated challenges by the significant development of novel methods and algorithms. In particular, network-based learning is providing promising results toward more personalized medicine. This short survey will describe three main interconnected trends identified to address these challenges and all with a firm root in network science: differential network analysis, network-based regularization, and causal discovery and inference. An overview of the main applications is provided, along with available software. Biomedical networks support more informed and interpretable statistical learning models from patients’ data, thus improving clinical decisions and supporting therapy optimization.

Marta B. Lopes, Susana Vinga

Chapter 4. Simultaneous Clustering of Multiple Gene Expression Datasets for Pattern Discovery

Background

Healthy cells run sophisticated genetic programmes in order to carry out their various biological processes such as cellular growth, cell division, stress response, and metabolism. The regulation of these genetic programmes is realised at different levels by controlling the production of the required types of large biomolecules such as RNAs, proteins, glycans, and lipids, with different amounts, at different times, and in different sub-cellular locations. Although all cells in an organism, such as skin cells, liver cells, bone cells, and neurons nominally have the same genomic material, they differ in shape and function because of the differences in the genetic programmes that they run.

Basel Abu-Jamous, Asoke K. Nandi

Chapter 5. Artificial Intelligence for Drug Development

Abstract

Drugs are treated as life-saving medicines against life-threatening diseases. However, drug developments pass through very complex and closely monitored phases to ensure the safety and efficacy of the intended purpose. The efforts are to keep highly toxic drugs from reaching even clinical trials. Even after the approval for drug distribution in the market, the drug’s post-marketing safety is analyzed by the number of reported Adverse Events (AEs). It requires the analysis and interpretation of massive data in all three stages namely pre-clinical, clinical and post-marketing stages. In this article, we explore the use of Artificial Intelligence (AI) in interpreting the huge data that is generated in the pre-clinical and clinical trials for safety purposes.

Muhammad Waqar Ashraf

6. Mathematical Bases for 2D Insect Trap Counts Modelling

Abstract

Pitfall trapping is a predominant sampling method in insect ecology, invasive species and agricultural pest management. Once samples are collected, their content is analyzed, different species are identified and counted and then used to provide reliable estimates of relative population abundance. Such estimates are essential for a variety of reasons, such as the general survey of insect diversity, detection of new insect invasions or simply for monitoring population levels. However, interpreting trap counts is a challenging task, since captures can depend on a variety of factors, such as environmental conditions, trap or survey design, insect movement behaviour, etc. Mathematical models provide an extremely useful description of how insects move in the field and in turn, can simulate the trapping process. In this chapter, we present the mathematical bases for 2D insect trap counts modelling, at the mean-field level using the diffusion equation and on an individual level using random walks. We reveal the intricacies of the trap counts dynamics, with details on how trap geometries and movement types can affect captures. We also describe the mathematical details for other trapping methods, such as baited trapping, where an attractant is used to lure the insects towards the trap location.

Danish A. Ahmed, Joseph D. Bailey, Sergei V. Petrovskii, Michael B. Bonsall

Medical Image Analysis

Frontmatter

7. Artificial Intelligence in Dermatology: A Case Study for Facial Skin Diseases

Abstract

The purpose of the first part of this chapter is to cover broadly the concept of using Artificial Intelligence (AI) in the field of dermatology. Afterward, it will mainly focus on facial skin diseases by covering some common AI-based approaches. The aim of this research application is the ability to identify certain pathologies by analyzing face images with present lesions through both AI and computer vision techniques. In particular, a special interest will be addressed to Machine-learning and Deep-learning approaches. In a case study, some key functionalities of a prototype software, developed by our research biometric group, will be presented.

Rola El-Saleh, Hazem Zein, Samer Chantaf, Amine Nait-ali

8. Medical Imaging Based Diagnosis Through Machine Learning and Data Analysis

Abstract

Machine learning techniques have played an essential role in computer-assisted medical image analysis. In this chapter, we will introduce several of our recent achievements with machine learning methods for feature extraction and representation, classification, dense prediction (segmentation and synthesis), and multi-modality analysis, across the pipeline of computer-assisted diagnosis (CAD). These methods consist of both traditional machine learning techniques and state-of-the-art deep learning based approaches. They were proposed to address pain points in the techniques, for example, similarity metric learning for better classification, 3D and sample-adaptive dense prediction models for segmentation and synthesis, and effective multi-modal imaging data fusion. These methods have been employed in different levels of medical imaging applications, such as medical image synthesis within and across imaging modalities, brain tumor segmentation, and mental disease classification. Common approaches used for related research topics are also briefly reviewed.

Jianjia Zhang, Yan Wang, Chen Zu, Biting Yu, Lei Wang, Luping Zhou

9. EfficientNet-Based Convolutional Neural Networks for Tuberculosis Classification

Abstract

Tuberculosis (TB) is an infectious disease that remained as a major health threat in the world. The computer-aided diagnosis (CAD) system for TB is one of the automated methods in early diagnosis and treatment, particularly used in developing countries. Literature survey shows that many methods exist based on machine learning for TB classification using X-ray images. Recently, deep learning approaches have been used instead of machine learning in many applications. This is mainly due to the reason that deep learning can learn optimal features from the raw dataset implicitly and obtains better performances. Due to the lack of X-ray image TB datasets, there are a small number of works on deep learning addressing the image-based classification of TB. In addition, the existing works can only classify X-ray images of a patient as TB or Healthy. This work presents a detailed investigation and analysis of 26 pretrained convolutional neural network (CNN) models using a recently released and large public database of TB X-ray. The proposed models have the capability to classify X-ray of a patient as TB, Healthy, or Sick but non-TB. Various visualization methods are adopted to show the optimal features learnt by the pretrained CNN models. Most of the pretrained CNN models achieved above 99% accuracy and less than 0.005 loss with 15 epochs during the training. All 7 different types of EfficientNet (ENet)-based CNN models performed better in comparison to other models in terms of accuracy, average and macro precision, recall, F1 score. Moreover, the proposed ENet-based CNN models performed better than other existing methods such as VGG16 and ResNet-50 for TB classification tasks. These results demonstrate that ENet-based models can be effectively used as a useful tool for TB classification.

Vinayakumar Ravi, Harini Narasimhan, Tuan D. Pham

Chapter 10. AI in the Detection and Analysis of Colorectal Lesions Using Colonoscopy

Abstract

Figure 10.1 illustrates the organization of this chapter. We begin with a brief review of colon anatomy and an overview of general information on colorectal cancers (CRCs). In Sect. 1.2, we introduce the details of colonoscopy, the most important tool for the screening, diagnosis, and therapy of CRCs.

Zhe Guo, Xin Zhu, Daiki Nemoto, Kazunori Togashi

11. Deep Learning-Driven Models for Endoscopic Image Analysis

Abstract

The advent of video endoscopy has led to an increased interest in the development of computer-aided diagnosis (CAD) approaches. Many of these focus on the use of deep learning methods as a means of automatically identifying abnormalities during endoscopy to lessen the workload on doctors. In this chapter, we take two tasks in endoscopic image analysis as examples, to survey the state of the art, recent advances, and future directions of CAD applications, especially with regard to deep learning models. We introduce the fundamentals of deep learning-driven methods and elaborate on their success in areas such as endoscopic image classification, detection of abnormal regions, and lesion boundary segmentation.

Xiao Jia, Xiaohan Xing, Yixuan Yuan, Max Q.-H Meng

Physiology

Frontmatter

Chapter 12. A Dynamic Evaluation Mechanism of Human Upper Limb Muscle Forces

Abstract

Dynamic evaluation mechanisms of the human upper limb are of great value for research and applications in upper limb rehabilitation, especially for the development of robotic upper limb rehabilitation systems. This paper proposes a muscle force prediction method based on the Hill muscle model. The proposed approach, which combines sEMG signals and kinematic data, provides a deep understanding of the dynamic motion mechanisms and parameters that characterize the upper limbs of the human body. The study provides a theoretical benchmark for the evaluation of rehabilitation training practices and for improved designs of upper limb rehabilitation robots that are used for upper limb neuro-rehabilitation. Specifically, the system collected motion data and sEMG signals from the upper limbs of the human body through a high-speed infrared motion capture system and skin sEMG sensors. By applying human kinematics and dynamics theories, real-time joint angle and torque information was obtained and imported into OpenSim. This platform can simulate the real-time muscle force values produced by the upper limbs during movements. The myoelectric signals were first filtered to remove noise, and an exponential model was then used to obtain the muscle activation. These data were then entered into the Hill-type prediction model to determine an individual’s muscle forces. In this paper, grasping movements commonly used in everyday situations were taken as a testing case. The results of the experiments showed that an individual’s muscle forces can be predicted using a Hill-type model. The results are consistent with those from simulated muscle force models and can reflect the real forces experienced during upper limb exercises.

Qing Tao, Zhaobo Li, Quanbao Lai, Shoudong Wang, Lili Liu, Jinsheng Kang

13. Resting-State EEG Sex Classification Using Selected Brain Connectivity Representation

Abstract

Electroencephalography (EEG) is a widely used non-invasive technique to measure multi-channel potentials that reflect the electrical activity of the brain. Over the last few decades, EEG analysis has been an intensively explored research topic due to its potentials in being applied to the diagnosis of neurological diseases, such as epilepsy, brain tumors, head injury, sleep disorders, and dementia [19]. Despite many advances made in recent years, EEG signal analysis remains a challenging task. In addition to being non-stationary, EEG signals often have high noise-to-information ratios, and they can be significantly affected by various artifacts, demonstrating characteristics that differ from signals generated by activities in the brain [21]. Common artifacts include eye movements, jaw tension, and muscle contractions. To make effective signal analysis even more challenging, EEG signals are highly individual-specific, and cross-subject pattern identification can be elusive.

Jean Li, Jeremiah D. Deng, Divya Adhia, Dirk De Ridder

Innovation in Medicine and Health

Frontmatter

14. Augmented Medicine: Changing Clinical Practice with Artificial Intelligence

Abstract

Smart medical technologies are augmenting clinical practice by offering the patient an increased autonomy and the clinician more advanced tools to predict, detect, monitor, and treat diseases. Augmented Medicine, a new framework of techniques that extends to clinical practice from the applied medical research aiming to introduce and improve tools, is rapidly gaining popularity. In this chapter, we will outline the principles of Augmented Medicine and its main applications in clinical practice, as well as future directions of this promising field.

Giovanni Briganti

15. Environmental Assessment Based on Health Information Using Artificial Intelligence

Abstract

A holistic care system which enables extensive medical care even outside the hospital brings significant benefits for health care. The application of novel communication and computation technologies is essential in order to accomplish such a system. In the presented chapter, a conceptual system is described which links environmental parameters measured by building automation and control systems with data from electronic health records. The system’s purpose is to provide medical personnel with interpreted data about possible adverse health effects of the indoor environment with respect to the patient’s health condition. Additionally, the patient receives real-time feedback about the environmental parameters and their potential health effects. The purpose of this feedback is to inspire behavior changes in the patient, which results in a more health-friendly environment. A special focus of the chapter lies on the analysis of possibly applicable artificial intelligence approaches for the estimation of the individual environmental risk factor. These are necessary because the system combines knowledge about the adverse health effect of environmental parameters and knowledge about health parameters for the environmental assessment. This knowledge is often incomplete, ambiguous, and is linked to uncertainty, which makes the interpretation of the raw data non-trivial and would overstrain the occupant as well as the medical personnel.

Jannik Fleßner, Johannes Hurka, Melina Frenken

16. Correction to: Advances in Artificial Intelligence, Computation, and Data Science

Tuan D. Pham, Hong Yan, Muhammad W. Ashraf, Folke Sjöberg

Backmatter

Title: Advances in Artificial Intelligence, Computation, and Data Science
Editors: Tuan D. Pham
Hong Yan
Muhammad W. Ashraf
Folke Sjöberg
Publisher: Springer International Publishing
Electronic ISBN: 978-3-030-69951-2
Print ISBN: 978-3-030-69950-5
DOI: https://doi.org/10.1007/978-3-030-69951-2