Demystifying Big Data, Machine Learning, and Deep Learning for Healthcare Analytics

Front Cover
Demystifying Big Data, Machine Learning, and Deep Learning for Healthcare Analytics
Copyright
Dedication
Contents
Contributors
Editors biography
Foreword
Preface
Overview
Section 1: Big data in healthcare analytics
	Chapter 1: Foundations of healthcare informatics
		1.1. Introduction
		1.2. Goals of healthcare informatics
		1.3. Focus of healthcare informatics
		1.4. Applications of healthcare informatics
		1.5. Medical information
		1.6. Clinical decision support systems
		1.7. Developing clinical decision support systems
			1.7.1. Traditional systems
			1.7.2. Evidence-based medicine
			1.7.3. Artificial intelligence and statistical inference-based approaches
		1.8. Healthcare information management
		1.9. Control flow
		1.10. Other perspectives
		1.11. Conclusion
		References
	Chapter 2: Smart healthcare systems using big data
		2.1. Introduction
			2.1.1. Background and driving forces
		2.2. Big data analytics in healthcare
			2.2.1. Disease prediction
			2.2.2. Electronic health records
			2.2.3. Real-time monitoring
			2.2.4. Medical strategic planning
			2.2.5. Telemedicine
			2.2.6. Drug suggestions
			2.2.7. Medical imaging
		2.3. Related work
		2.4. Big data for biomedicine
		2.5. Proposed solutions for smart healthcare model
		2.6. Role of sensor technology for eHealth
		2.7. Major applications and challenges
		2.8. Conclusion and future scope
		References
	Chapter 3: Big data-based frameworks for healthcare systems
		3.1. Introduction
		3.2. The role of big data in healthcare systems and industry
		3.3. Big data frameworks for healthcare systems
		3.4. Overview of big data techniques and technologies supporting healthcare systems
			3.4.1. Cloud computing and architecture
			3.4.2. Fog computing and architecture
			3.4.3. Internet of things (IoT)
			3.4.4. Internet of medical things (IoMT)
			3.4.5. Machine learning (ML)
			3.4.6. Deep learning
			3.4.7. Intelligent computational techniques and data mining
		3.5. Overview of big data platform and tools for healthcare systems
			3.5.1. Hadoop architecture
			3.5.2. Apache hadoop
			3.5.3. Apache spark
			3.5.4. Apache storm
		3.6. Proposed big data-based conceptual framework for healthcare systems
			3.6.1. Proposed system functionalities
				3.6.1.1. Data sources
				3.6.1.2. Patient healthcare-related data
				3.6.1.3. Cloud and fog computing components
				3.6.1.4. Big data analytics methods, techniques, and platform tools
				3.6.1.5. Patient healthcare monitoring and recommendation system
				6.1.6. Healthcare research and knowledge infrastructure
		3.7. Conclusion
		References
	Chapter 4: Predictive analysis and modeling in healthcare systems
		4.1. Introduction
		4.2. Process configuration and modeling in healthcare systems
		4.3. Basic techniques of process modeling and prediction
			4.3.1. Process discovery
			4.3.2. Enhancement
		4.4. Event log
			4.4.1. Event and attributes
			4.4.2. Case, trace, and event log
			4.4.3. Structure of an event log
		4.5. Control perspective of hospital process using various modeling notations
			4.5.1. Transition systems
			4.5.2. Petri net
			4.5.3. Workflow nets
			4.5.4. Yet another workflow language (YAWL)
			4.5.5. Business process modeling notation (BPMN)
			4.5.6. Event-driven process chains (EPC)
			4.5.7. Causal nets
		4.6. Predictive modeling control flow of a process using fuzzy miner
			4.6.1. Hospital process
			4.6.2. Hospital treatment process
		4.7. Open research problems
		4.8. Conclusion
		References
	Chapter 5: Challenges and opportunities of big data integration in patient-centric healthcare analytics using mobile networks
		5.1. Introduction
		5.2. Elderly health monitoring using big data
			5.2.1. eHealth
			5.2.2. General health issues in the elderly
		5.3. Personalized monitoring and support platform (MONISAN)
			5.3.1. Proposed development
		5.4. Patient-centric healthcare provider using big data
			5.4.1. Resource allocation in mobile networks using big data analytics: A survey
			5.4.2. Healthcare analytics: A survey
		5.5. Patient-centric optimization model
			5.5.1. Structure model
			5.5.2. Classification using naïve Bayesian
			5.5.3. Reduction of data
			5.5.4. Generalization of data
			5.5.5. The naïve Bayesian formulation techniques used to calculate patient priority using MILP
			5.5.6. Formulation of problem
		5.6. The WSRMAX approach-based MILP formulation
			5.6.1. The optimization techniques used before providing priority to patients
		5.7. MILP formulation-probability fairness approach
			5.7.1. The optimization techniques used before providing priority to patients
			5.7.2. After patients prioritization
				5.7.2.1. Receiving power calculation
		5.8. Heuristic approach
		5.9. Results and discussion
			5.9.1. The WSRMAX approach-based MILP and heuristic formulation
				5.9.1.1. The optimization techniques used before providing priority to patients
				5.9.1.2. After patient prioritization
			5.9.2. Probability fairness approach
				5.9.2.1. The optimization techniques used before providing priority to patients
				5.9.2.2. After patient prioritization
		5.10. Future directions
			5.10.1. Choice of decision-making platform
			5.10.2. Ranking features and selecting the most optimized feature
			5.10.3. Integration with 5G
			5.10.4. Infrastructure sharing
			5.10.5. Wireless drug injection
		5.11. Conclusion
		References
	Chapter 6: Emergence of decision support systems in healthcare
		6.1. Introduction
			6.1.1. Overview
			6.1.2. Need for CDSS
			6.1.3. Types of CDSS
			6.1.4. Effectiveness and applications of CDSS
		6.2. Transformation in healthcare systems
			6.2.1. Adoption of CDSS
			6.2.2. Key findings
			6.2.3. Enterprise-level adaptation
			6.2.4. Health IT infrastructure
		6.3. CDS-based technologies
			6.3.1. Supervised learning techniques
				6.3.1.1. Decision tree
				6.3.1.2. Logistic regression
				6.3.1.3. Neural networks
			6.3.2. Unsupervised learning techniques
			6.3.3. Disease diagnosis techniques
				6.3.3.1. Domain selection
				6.3.3.2. Knowledge base-construction
				6.3.3.3. Algorithms and user interface
			6.3.4. CDS-related issues
		6.4. Clinical data-driven society
			6.4.1. Information extraction
			6.4.2. CDS today and tomorrow
		6.5. Future of decision support system
		6.6. Example: Decision support system
			6.6.1. CDSS for liver disorder identification
		6.7. Conclusion
		References
Section 2: Machine learning and deep learning for healthcare
	Chapter 7: A comprehensive review on deep learning techniques for a BCI-based communication system
		7.1. Introduction
			7.1.1. Brain signals
				7.1.1.1. Brain computer interface
				7.1.1.2. Electric potential source for BCI
				7.1.1.3. Evoked potential
				7.1.1.4. Event-related potential (ERP)
		7.2. Communication system for paralytic people
			7.2.1. Oculography-based control systems
			7.2.2. Morse code-based assistive tool
			7.2.3. Sensor-based systems
			7.2.4. EEG-based systems
		7.3. Acquisition system
			7.3.1. Benchmark datasets
		7.4. Machine learning techniques in EEG signal processing
			7.4.1. Support vector machine
			7.4.2. k-NN
			7.4.3. Logistic regression
			7.4.4. Naïve Bayes
		7.5. Deep learning techniques in EEG signal processing
			7.5.1. Deep learning models
				7.5.1.1. Supervised deep learning
				7.5.1.2. Convolutional neural network (CNN)
				7.5.1.3. RNN
				7.5.1.4. Unsupervised deep learning
				7.5.1.5. Autoencoder
				7.5.1.6. Semisupervised deep learning
				7.5.1.7. Deep belief network
			7.5.2. Deep learning in feature extraction
			7.5.3. Deep learning for classification
		7.6. Performance metrics
		7.7. Inferences
		7.8. Research challenges and opportunities
			7.8.1. Using multivariate system
			7.8.2. The dimensionality of the data
			7.8.3. Artifacts
			7.8.4. Unexplored areas
		7.9. Future scope
		7.10. Conclusion
		Acknowledgments
		References
	Chapter 8: Clinical diagnostic systems based on machine learning and deep learning
		8.1. Introduction
		8.2. Literature review and discussion
			8.2.1. Major findings in problem domain
		8.3. Applications of machine learning and deep learning in healthcare systems
			8.3.1. Heart disease diagnosis
			8.3.2. Predicting diabetes
			8.3.3. Prediction of liver disease
			8.3.4. Robotic surgery
			8.3.5. Cancer detection and prediction
			8.3.6. Personalized treatment
			8.3.7. Drug discovery
			8.3.8. Smart EHR
		8.4. Proposed methodology
			8.4.1. Intraabdominal ultrasound image acquisition
			8.4.2. Ultrasound image enhancement
			8.4.3. Segmentation of the RoI image
			8.4.4. Intraabdominal organ identification using deep neural network
			8.4.5. Feature extraction
			8.4.6. Abnormality identification and categorization
		8.5. Results and discussion
		8.6. Future scope and perceptive
		8.7. Conclusion
		References
	Chapter 9: An improved time-frequency method for efficient diagnosis of cardiac arrhythmias
		9.1. Introduction
		9.2. Methods
			9.2.1. Dual tree wavelet transform
			9.2.2. Support vector machines (SVMs)
			9.2.3. PSO technique
		9.3. Proposed methodology
			9.3.1. Database
			9.3.2. Denoising
			9.3.3. QRS wave localization and windowing
			9.3.4. Input representation
			9.3.5. Feature classification
			9.3.6. Performance metrics
		9.4. Experiments and simulation performance
			9.4.1. Evaluation in patient-specific scheme
			9.4.2. Advantages of proposed method
		9.5. Conclusion and future scope
		References
	Chapter 10: Local plastic surgery-based face recognition using convolutional neural networks
		10.1. Introduction
		10.2. Overview of convolutional neural network
			10.2.1. Convolutional layer
			10.2.2. Pooling layers
			10.2.3. Fully connected layers
			10.2.4. Activation functions
			10.2.5. CNN architectures
				10.2.5.1. LeNet
				10.2.5.2. AlexNet
				10.2.5.3. ZFNet
				10.2.5.4. VGG
				10.2.5.5. GoogLeNet
				10.2.5.6. ResNet
		10.3. Literature survey
		10.4. Design of deep learning architecture for local plastic surgery-based face recognition
			10.4.1. Proposed CNN model
				10.4.1.1. Convolution layer
				10.4.1.2. Pooling layer
				10.4.1.3. Fully connected layer
				10.4.1.4. Parameter tuning
		10.5. Experimental setup
		10.6. Database description
		10.7. Results
		10.8. Conclusion and future scope
		References
	Chapter 11: Machine learning algorithms for prediction of heart disease
		11.1. Introduction
			11.1.1. Introduction to ML
			11.1.2. Types of ML
				11.1.2.1. Supervised learning
				11.1.2.2. Unsupervised learning
				11.1.2.3. Semisupervised learning
				11.1.2.4. Reinforcement learning
		11.2. Literature review
		11.3. ML workflow
			11.3.1. Data collection
			11.3.2. Cleaning and preprocessing
			11.3.3. Feature selection
			11.3.4. Model selection
			11.3.5. Training and evaluation
		11.4. Experimental setup
		11.5. Supervised ML algorithms
			11.5.1. Support vector machine
			11.5.2. Logistic regression
			11.5.3. Decision tree
			11.5.4. Naive Bayes classifier
		11.6. Ensemble ML models
			11.6.1. Majority voting
			11.6.2. Weighted average voting
			11.6.3. Bagging
			11.6.4. Gradient boosting
		11.7. Results and discussion
			11.7.1. Visualization of performance metrics of base learners
			11.7.2. Visualization of performance metrics of ensemble learners
		11.8. Summary
		Acknowledgments
		References
	Chapter 12: Convolutional Siamese networks for one-shot malaria parasite recognition in microscopic images
		12.1. Introduction
		12.2. Related works
			12.2.1. State-of-the-art methods for one-shot learning
			12.2.2. Siamese network for face recognition and verification
			12.2.3. Siamese network for scene detection and object tracking
			12.2.4. Siamese network for two-stage learning and recognition
			12.2.5. Siamese network for medical applications
			12.2.6. Siamese network for visual tracking and object tracking
			12.2.7. Siamese network for natural language processing
		12.3. Materials and methods
		12.4. Proposed methodology
			12.4.1. Siamese neural architecture
			12.4.2. Training, parameter tuning, and evaluation
		12.5. Results and discussions
		12.6. Conclusions
		References
	Chapter 13: Kidney disease prediction using a machine learning approach: A comparative and comprehensive analysis
		13.1. Introduction
			13.1.1. Causes of chronic kidney disease
			13.1.2. Detection of chronic kidney disease
			13.1.3. Treatments for chronic kidney disease
		13.2. Machine learning importance in disease prediction
		13.3. ML models used in the study
			13.3.1. KNN classifier
			13.3.2. Logistic regression
			13.3.3. Support vector machine
			13.3.4. Random forest
			13.3.5. Naïve Bayes
			13.3.6. Artificial neural network
			13.3.7. AdaBoost
		13.4. Results and discussion
			13.4.1. Quality measurement
				13.4.1.1. Information gain
				13.4.1.2. Gain ratio
				13.4.1.3. Gini Index
				13.4.1.4. Chi-squared distribution
				13.4.1.5. FCBF
			13.4.2. Evaluation techniques
				13.4.2.1. Confusion matrix
				13.4.2.2. Receiver operating characteristic (ROC) curve
			13.4.3. Dataset description
			13.4.4. Model configurations
			13.4.5. Result analysis with information gain
			13.4.6. Result analysis with information gain ratio
			13.4.7. Result analysis with Gini Index
			13.4.8. Result analysis with chi-square
		13.5. Conclusion
		References
Index
Back Cover