Artificial Intelligence for Healthcare Applications and Management

Artificial Intelligence for Healthcare Applications and Management introduces application domains of various AI algorithms across healthcare management. Instead of discussing AI first and then exploring its applications in healthcare afterward, the authors attack the problems in context directly, in order to accelerate the path of an interested reader toward building industrial-strength healthcare applications. Readers will be introduced to a wide spectrum of AI applications supporting all stages of patient flow in a healthcare facility. The authors explain how AI supports patients throughout a healthcare facility, including diagnosis and treatment recommendations needed to get patients from the point of admission to the point of discharge while maintaining quality, patient safety, and patient/provider satisfaction. AI methods are expected to decrease the burden on physicians, improve the quality of patient care, and decrease overall treatment costs. Current conditions affected by COVID-19 pose new challenges for healthcare management and learning how to apply AI will be important for a broad spectrum of students and mature professionals working in medical informatics. This book focuses on predictive analytics, health text processing, data aggregation, management of patients, and other fields which have all turned out to be bottlenecks for the efficient management of coronavirus patients. Presents an in-depth exploration of how AI algorithms embedded in scheduling, prediction, automated support, personalization, and diagnostics can improve the efficiency of patient treatment Investigates explainable AI, including explainable decision support and machine learning, from limited data to back-up clinical decisions, and data analysis Offers hands-on skills to computer science and medical informatics students to aid them in designing intelligent systems for healthcare Informs a broad, multidisciplinary audience about a multitude of applications of machine learning and linguistics across various healthcare fields Introduces medical discourse analysis for a high-level representation of health texts Front Matter Copyright Contributors Introduction The issues of ML in medicine this book is solving AI for diagnosis and treatment Health discourse Acknowledgments Supplementary data sets References Multi-case-based reasoning by syntactic-semantic alignment and discourse analysis Introduction Multi-case-based reasoning in the medical field Mixed illness description Probabilistic ontology Mapping a patient record to identified cases Learning semantic similarity Discourse analysis and discourse trees Alignment of linguistic graphs Abstract meaning representation Aligning AMR Alignment algorithm Case-based reasoning in health Mining cases from health forum threads Discourse disentanglement Building a repository of labeled cases and diagnoses An example of navigating an extended discourse tree for three documents Constructing extended DTs System architecture Evaluation Datasets Evaluation of text matching Overall assessment of the Symptom Checker engine Diagnosing forum data Related work Conclusions Supplementary data sets References Obtaining supported decision trees from text for health system applications Introduction Supported decision trees for an expert system Obtaining supported decision trees from text From a discourse tree to its supported decision tree System architecture of construction of a supported decision tree Evaluation Decision trees in health Defining decision tree as a supervised learning task Decision trees and COVID-19 Expert system for health management Basic expert systems and their values in health domain Backward chaining inference Expert system and health management Clinical use of expert systems Expert system lifecycle Learning ES rules Dynamics of ES usage Conclusions Supplementary data sets References Search and prevention of errors in medical databases Introduction Data entry errors when transferring information from the initial medical documentation to the studied database Analyzed databases Impossible/internally inconsistent data Externally inconsistent data Impossible/internally inconsistent data entry in B and S databases Specific type of the errors ``omitted data ́ ́ Errors in initial medical information Measurement errors ``bodyweight ́ ́ as an indicator of the quality of the initial information Algorithm Errors in EMR data User history of previous errors Physicians vs non-physicians Effect of practice location on weight error rates Error rates over time Error rates and user experience Error reduction Detection errors in datasets Alarm system in data entry process ``Follow-up summary ́ ́ as a method of error prevention ``Follow-up summary ́ ́ description ``Follow-up summary ́ ́ implementation ``Follow-up summary ́ ́ utilization ``Follow-up summary ́ ́ effectiveness Conclusions Supplementary data sets References Overcoming AI applications challenges in health: Decision system DINAR2 Introduction Problems of introducing medical AI applications Domain overfitting Terminology problems Cognitive bias Integration of AI into clinical practice Integrated decision support system at the regional consultative Center for Intensive Pediatrics (DINAR2) Idea and problems of the regional consultative Center for Intensive Pediatrics History of DINAR2 development Methods Considering data provided by a doctor as input for the operation of DINAR2 as fuzzy sets Construction of diagnostic rules An assessment of a patients state of severity A definition of the leading pathological syndrome Methods of stimulation of intellectual activity of a CIP consultant DINAR2s operation in extreme situations Freedom of a consultants actions within DINAR2 limits Organization of a database of local hospitals Stimulating doctors to improve DINAR2 DINAR2 efficiency Conclusions Supplementary data sets References Formulating critical questions to the user in the course of decision-making Introduction Reasoning patterns and formulating critical questions Automated building of reasoning chains Questions as relative complement of linguistic representations Generating text from AMR graph fragment Deriving critical questions via anti-unification Question-generation system architecture Chatbot implementation Data collection Evaluation Syntactic and semantic generalizations Semantic generalization Attribute-based generalization Building questions via generalization of instances Discussion and conclusions Supplementary data sets References Relying on discourse analysis to answer complex questions by neural machine reading comprehension Introduction Examples where discourse analysis is essential for MRC Discourse dataset Discourse parsing Incorporating syntax into model Attention mechanism for the sequence of tokens Enabling attention mechanism with syntactic features Including discourse structure into the model Pre-trained language models and their semantic extensions Encoding and alignment with BERT Direct similarity-based question answering Correcting an MRC answer System architecture Evaluation Discussion and conclusions Supplementary data sets References Machine reading between the lines (RBL) of medical complaints Introduction RBL, machine reading comprehension, and inference RBL and common sense RBL as generalization and web mining Patient repeats what he wants to say Reading deep between the lines RBL in storytelling Extracting RBL results from text Difficult RBL cases RBL in a dialogue Question formation and diversification System architecture Statistical model of RBL RBL and NLI NLI and semantic fragments Reinforcement learning approach Language models Storytelling discourse approach Evaluation Meaningfulness of generated RBLs Search recall improvement Discussions Conclusions Supplementary data sets References Discourse means for maintaining a proper rhetorical flow Introduction Medical dialogue systems Discourse tree of a dialogue Response selection Speech acts and communicative actions A dialogue with doubt Further extending the set of rhetorical relations toward dialogue Computing rhetorical relation of entailment Dialogue generation as language modeling Strategies for informative conversations Rhetorical agreement between questions and answers Discourse parsing of a dialogue Constructing a dialogue from text Building a dialogue based on a DT Constructing questions System architecture Evaluation Discussions and conclusions Supplementary data sets References Dialogue management based on forcing a user through a discourse tree of a text Introduction Keeping a learner focused on a text Navigating discourse tree in conversation The dialogue flow Managing user intents Handling epistemic states User intent recognizer Nearest neighbor-based learning for user intent recognition System architecture Evaluation Evaluation setting Assessment of navigation algorithm Related work Personalization in health chatbots Interaction in the mental space Persuasive dialogue Conclusions Supplementary data sets References Building medical ontologies relying on communicative discourse trees Introduction Ontology extraction from text Text mining Introducing discourse features Discourse-level support for ontology construction Issues associated with not using discourse information for ontology entry extraction Annotating events Informative and uninformative parts of text Informative and uninformative parts of an answer How a discourse tree indicates what to index and what not to index How rhetorical relations determine indexing rules Designing ontologies Systematized nomenclature of medicine-Clinical terms Relation extractor based on syntactic parsing Conceptualization process Neural dictionary manager Phrase aggregator Ontologies supporting reasoning Entity grid helps to extract relationships Validating ontology Specific ontology types in bioinformatics Spatial taxonomy Supporting search System architecture Evaluation Datasets Assessment of ontology consistency An assessment of search improvement due to ontology Conclusions Supplementary data sets References Explanation in medical decision support systems Introduction Models of machine learning explanation Interpretable models Black-box models Explanation based on comparison of the local case with the closest case with an alternative ML solution Finding the closest point to a local case A bi-directional adversarial meta-agent between user and ML system Meta-agent behavior Steps of the meta-agent Discussion Conclusions Supplementary data sets References Passive decision support for patient management Introduction Dr. Watson-type systems Principles of Dr. Watson-type systems Dr. Watson-type system formalization Patient management system (SAGe) Requirements and subsystems Information import Diagnostics Treatment effectiveness Treatment adequacy Discontinuation of observation Integral assessment of patients in the department Features of Dr. Watson-type system presented in SAGe Discovery of contradictions and omissions Attempts to direct physicians towards alternative solutions Encouragement and motivation Conclusions Supplementary data sets References Multimodal discourse trees for health management and security Introduction Forensic linguistics Extended discourse trees Victims right and state responsibility to investigate Discourse analysis of health and security-related scenarios Discourse of a reasonable doubt Discourse analysis of a scenario Multimodal discourse representation Multimodal discourse tree for a crime report Multimodal data sources and references between them Manipulation with discourse trees Extended discourse tree Mobile location data and COVID-19 Call detail records and COVID-19 Automatic number plate recognition Reasoning about a cause and effect of data records Representing causal links by R-C framework Reasoning with arguments extracted from text System architecture Evaluation Discussions and conclusions Supplementary data sets References Improving open domain content generation by text mining and alignment Introduction Content generation in health care Content generation for personalization Natural language generation in intensive care Processing raw natural language generation results Alignment of raw and true content Fact-checking of deep learning generation Personalized drug recommendation Discourse structure deviation of the corrected content System architecture Deep learning subsystem Raw content correction Probabilistic text merging Graph-based fact-checking Entity substitution Evaluation Discussions Conclusions Supplementary data sets References