Artificial Intelligence in Precision Health: From Concept to Applications

Front matter
Copyright
Dedication
Contributors
Editor's biography
Preface
Interpretable artificial intelligence: Closing the adoption gap in healthcare
	Artificial intelligence in healthcare
	Why do we need interpretable intelligent systems in healthcare?
		Right to explanation and the regulatory landscape
		Medicine as a quest for ``why´´
		The need for a culture of AI-assisted healthcare
		Adoption in clinical decision-making
		Relevance in the marketplace
	What does interpretability mean?
	How to realize interpretability in intelligent systems?
		Achieving interpretability by design
			Case study: Predicting hemodynamic shock from thermal images using machine learning
		Achieving interpretability through inherently transparent models
			Linear and logistic regression models
			Decision tree models (Quinlan, 1986)
			Generalized additive models and partial dependence plots
		Achieving model interpretability through post hoc methods
			Feature importance
			Boruta
			Shapley values (SHAP)
			Surrogate trees
			Locally interpretable model-agnostic explanations (LIME)
		Achieving interpretability through graphical models
		Achieving interpretability in deep neural networks
		Taxonomy of interpretable deep learning methods
		Backpropagation-based methods
			Deconvolution
			Saliency maps
			Guided backpropagation
			Integrated gradients
			SmoothGrad
			Layer-wise relevance potential (LRP)
			DeepLIFT
		Perturbation-based techniques
			Lime
			Occlusion sensitivity
		Activation maximization
			Class model visualization (CMV)
			Grad-CAM and Grad-CAM++
	Summary and road map for the future
	Acknowledgments
	References
Artificial intelligence methods in computer-aided diagnostic tools and decision support analytics for clinical ...
	Introduction
	Artificial intelligence methods and applications
		Genetic algorithm
			Applications of genetic algorithms
		Support vector machines
			Applications of support vector machines
		Artificial neural networks and deep learning
			Application of artificial neural networks and deep learning
		Decision trees
			Case study: Predicting heart diseases
		k-Nearest neighbors
			Case study: Finding similar patients
		k-Means
			Case study: Clustering heart disease data
			Case study: Correlating gene expression to colorectal cancer outcomes
		Natural language processing
			Applications of natural language processing
	From concepts to applications
		Application: HINGE-A radiation oncology analytics portal
	Conclusion
	References
Deep learning in precision medicine
	Introduction to deep learning
	Hardware and software requirements for deep learning
		Hardware-GPU cards
		Software-Deep learning package
	ANN, CNN, and deep learning concepts
		Concepts
	How deep learning transforms the study of human disease?
		Deep learning and clinical decision-making
		Deep learning and patient categorization and precision/predictive medicine
		Deep learning to study the fundamental biological processes underlying human disease
		The impact of deep learning in treating disease and developing new and personalized treatments
	An example of deep learning implementation in medicine
		Binary class definition
		Multiclass definition
		Encoder-decoder architecture
		End to end example
			Exploring the dataset and data preparation
			Data preprocessing
			Model implementation
				Setting up the environments and dependencies
				Building the blocks of the network
				Building the model
			Training the model
			Model predictions
	Conclusion and future directions
	Acknowledgments
	References
Machine learning systems and precision medicine: A conceptual and experimental approach to single individual s ...
	Introduction: Personalized medicine and precision medicine
	First case study: Self-organizing maps (SOMs) and the case of quality-of-life scales
		The SOM algorithm
		Clinical application
	Second case study: Pick-and-squash tracking (PST) algorithm to cluster patients with and without Barrett disease
		The PST algorithm
		Clinical application
	Third case study: Clustering of patients with and without myocardial infarction by means of auto-contractive map (auto-CM)
		Auto-CM neural algorithm
		Clinical application
	Fourth case study: Use of several different machine learning systems to classify the single individual allowing degree of c ...
		General philosophy of the approach
		Is there any solution to this problem?
		Clinical application
	Discussion
	Conclusions and future direction
	References
	Further reading
Machine learning in digital health, recent trends, and ongoing challenges
	Introduction
	Training and testing: The machine learning pipeline
	Machine learning algorithms
		Generative models
		Discriminative models
		Toolkits
	Machine learning in action: Exemplary tasks and case studies
		Snore sound detection
		Abnormal heart sound classification
	Challenges and future work directions
		Increased explainability
		Deployment of AI in mobile and embedded technologies
		Data sparsity
	Conclusion
	Acknowledgments
	References
Data mining to transform clinical and translational research findings into precision health
	Introduction
	Data mining strategies and techniques in clinical and translational research
		Data mining applications in health care
		Data mining in clinical and translational research
		Data mining strategies and techniques
		Machine learning applications
		Data mining research and infrastructure
	Translating data mining to advance genomics in disease risk
		Healthy people
		Polygenic risk scores
		Translation initiatives to advance genomics in precision health
	Role of clinical research data warehousing in ``big data´´ science
		Data format
		Data sources
		Data model to knowledge model
	Integration of multiple data sources to advance precision health
		Environmental
		Behavioral
		Imaging
	Conclusion
	Future direction
	References
	Further reading
Predictive models in precision medicine
	Introduction
	Predictive analysis
	Predictive modeling
		Predictive models
		Precision medicine
		How predictive modeling works in precision medicine
			Generalized linear models
			Decision trees
			Artificial neural networks
			Support vector machines
			Expert systems
			Naïve Bayes
			K-nearest neighbor
			Random forest
			Logistic regression
			Time series analysis
			Fuzzy logic
			Other methods and medical areas of use
			Real-time applications
	Conclusions and future directions
	References
	Further reading
Deep neural networks for phenotype prediction in rare diseases: Inclusion body myositis: A case study
	Introduction
	Case study-inclusion body myositis
	Efficacy of the method
	Conclusion
	Acknowledgments
	References
Artificial intelligence for management of patients with intracranial neoplasms
	Introduction
	Diagnosis
		ML for medical imaging
			ML for image segmentation
			Virtual biopsy with ML
		AI and histopathology
	AI for treatment
		AI and decision-making
		AI in neurosurgery
			AI for surgery simulation
			AI for intraoperative assistance
		AI in postoperative care
		AI for radiation therapy
	AI for prognosis
	Future challenges and directions
	Conclusions
	References
Artificial intelligence to aid the detection of mood disorders
	Introduction
	The case for AI-based objective diagnostic markers
	Machine learning: A brief introduction
	Data relating to mood disorders
		Physiological data
		Digital-trace information
		Audio-visual information
	Software platforms and smartphone applications
	AI in action: Depression and bipolar disorder detection
		Depression detection
		Bipolar disorder detection
	Challenges and future work directions
	Conclusion
	Acknowledgment
	References
Use of artificial intelligence in Alzheimers disease detection
	Introduction
	Artificial intelligence techniques in Alzheimers disease detection
		Artificial neural networks
		K-nearest neighbor (k-NN)
		Support vector machines (SVM)
		Random forest
		Ensemble classifiers
		Deep neural networks
		Convolutional neural networks
	Why artificial intelligence is important for AD
	Conclusions and future directions
	References
Artificial intelligence to predict atheroma plaque vulnerability
	Introduction
	Atheroma plaque vulnerability: Case of study
		Modeling of the atherosclerotic coronary artery
			Idealized geometry
			Parameters studied
			Mesh
			Material properties
			Boundary conditions and loads
		Results
		Statistical analysis
		Vulnerability study
	Machine learning techniques (MLT) as a helpful tool toward determination of plaque vulnerability
		Data acquisition and preprocessing
		Mathematical methods for regression
			Artificial neural network (ANN)
			Support vector machine (SVM)
			Classical linear regression
		Performance and accuracy of the regressor
		How does the decision support system work?
		Results of the vulnerability prediction
	Discussion
	Conclusions and future directions
	Acknowledgments
	References
Artificial intelligence in cardiovascular medicine: Applications in the diagnosis of infarction and prognosis ...
	Introduction
	Summary of the main artificial intelligence algorithms
		Artificial neural networks and deep learning
		Decision trees
		Support vector machines
	Application of artificial intelligence to the diagnosis of acute coronary syndromes and acute myocardial infarction
		Historical aspects
		Context of application
	Artificial intelligence applied to the prognosis of heart failure
		Works based on clinical and laboratory data
		Works that included biomarkers
		Works based on echocardiography or effort tests
		Telemonitoring-based works
	Conclusions and future directions
	References
Artificial intelligence-based decision support systems for diabetes
	Introduction
	Diabetes management
		T1D treatment
	Blood glucose prediction
	Prediction of glycemic episodes
	Insulin bolus calculators and advisory systems
	Risk and patient stratification
	Commercial systems
	Conclusions
	Future directions
	Acknowledgments
	References
Clinical decision support systems to improve the diagnosis and management of respiratory diseases
	Introduction
	A brief review of the machine learning methods used in respiratory care
		Logistic regression
		K-nearest neighbor (KNN)
		Decision tree (DTREE)
		Artificial neural networks (ANNs)
		Support vector machines
		Random forest (RF)
		AdaBoost
		Performance evaluation and hypothesis test
	Brief introduction to the methods of pulmonary function analysis
		Spirometry
		Forced oscillation technique
	Artificial intelligence/machine learning methods to improve the pulmonary function analysis
		Spirometry
			The first studies in the 1980s
			Studies performed in the 2000s
			Studies performed in the 2010s
		Forced oscillation technique (FOT)
		Miscellaneous pulmonary function methods
		Telemedicine
		Examples of commercial systems
	Possible future directions
		Big data analytics
		Interactive machine learning
		Deep learning
	Conclusions and future directions
	References
Artificial intelligence in neuro, head, and neck surgery
	Introduction
	Artificial intelligence in head and neck surgery
		State-of-art
		Precision systems used in otorhinolaryngology
		Recent studies
			Otology
			Rhinology and infections
			Oral and laryngology
			Reconstructive surgeries of head and neck
			Oncology
			Education
	Artificial intelligence in neurosurgery
		Recent studies
			Robotic surgery
			Neurovascular surgery
			Neurooncology
			Trauma
			Spinal surgery
			Neuroimaging
		Precision systems used in routine practice
	Conclusions and future directions
	References
	Further reading
Use of artificial intelligence in emergency medicine
	Medical informatics on emergency medicine
	Artificial intelligence
	Artificial intelligence and emergency medicine
	Artificial intelligence studies in emergency medicine
		Triage
		Cardiac arrest
		Cardiovascular events diagnosis
		Stroke
		Sepsis
		Prediction of admission and visits
	Commercial precision systems used in emergency care
	Conclusion and future aspects
	References
	Further reading
Use of artificial intelligence in infectious diseases
	Preamble on infectious diseases
	Artificial intelligence in health care
	The utilization of AI in infectious diseases
		Improved diagnosis and blocking transmission
			Diagnosis
			Epidemiology and transmission
		Treatments and antimicrobial drug resistance
	Improving the process
		On the technical aspects
		The potential of extreme value theory
		Basics on the concept of extreme values
		On the design of data collection
		On the integration of AI in health-care institutions
	Conclusions and future perspectives
	Acknowledgments
	References
Artificial intelligence techniques applied to patient care and monitoring
	Introduction
	Patient care scenarios
	Artificial intelligence approaches for health care
	Data gathering and feature extraction
	Data analysis
	Feedback generation
		Patient safety through smart notifications
		Inferring context using artificial intelligence
	Challenges and future directions
	References
Use of artificial intelligence in precision nutrition and fitness
	Introduction
		The importance of nutrition and fitness for health and well-being
		What is precision medicine: Concepts and historical aspects
		What is artificial intelligence: Concepts and historical aspects related to its use in nutrition and fitness
			Fuzzy logic
			Artificial neural networks
			Evolutionary computing
		What is precision nutrition and precision fitness: Clarifying the concepts
		How AI could help with precision nutrition
			Decision-making algorithm for nutritional meal planning/dietary menu planning
			Artificial intelligence-based diet and supplements
			AI used in genetic tests for precision nutrition and fitness
			Artificial intelligence approach to nutritional meal planning for cancer
			Artificial intelligence approach to nutritional meal planning for cardiovascular diseases
			Artificial intelligence approach to nutritional meal planning for obesity (weight management/loss)
			Artificial intelligence approach to nutritional meal planning for T2D patients
			Artificial intelligence-based nutrition and fitness support systems and apps (free and commercial)
	How AI could help with precision fitness
	Challenges and future perspectives
	References
Artificial intelligence in precision health: Systems in practice
	Introduction
		Concept of precision health in the era of artificial intelligence
	History and approaches of artificial intelligence in precision health
	Applications of machine-learning approaches in precision health
		Case-based reasoning: k-nearest neighbor
		Case-based reasoning (CBR): Other techniques
		K-means clustering
		Logistic regression, linear discriminant analysis, principal components analysis
		Support vector machines
		Decision trees
		Random forests
		Bayesian networks and Naïve Bayesian Classifiers (NBC)
		Artificial neural networks (ANN)
		Deep learning
		Genetic algorithms
		Artificial immune systems
		Ensembles
		Repositories
	Systems in place: AI-based commercial decision support systems in precision health practice
		IBM Watson (www.ibm.com/watson)
		Isabel Healthcare (www.isabelhealthcare.com)
		Symptomate (www.symptomate.com)
		GeNIe and SMILE (www.bayesfusion.com/genie/)
	Other differential diagnosis generators
		Crowdsourcing
	Other intelligent tools of interest
		Sophia genetics (www.sophiagenetics.com)
		Genetic therapies-Deep genomics (www.deepgenomics.com)
		Genomic and artificial intelligence solutions-BioRealm (www.biorealm.ai)
		DeepVariant and DeepMind by Google (www.ai.google/healthcare; www.deepmind.com; www.cloud.google.com; www.github.com/google ...
		Emedgene platform-Emedgene (www.emedgene.com)
		Personalized medical decision-making: Flow Health (www.flowhealth.com)
		Panorama (NIPT)-Natera (www.natera.com)
		Nebula Genomics (www.nebula.org)
		Tempus (www.tempus.com)
		Cognitive mobile health care: Pathway Genomics and Apple Health Kit (www.pathway.com)
		Helix (www.helix.com)
	Conclusions and future directions
	References
Index