M.J. Fernandez-Prieto

Electronic Health Records (EHRs) Standards and the Semantic Edge: A Case Study of Visualising Clinical Information from EHRs

Exchanging medical documents over healthcare networks is becoming a reality. This increases the need to effectively manage the growing amount of information for a single patient. Therefore, there is a current need to visualise Electronic Health Records (EHRs) in a way that assist physicians with clinical tasks and medical decision-making. The new methods to visualise clinical information from EHRs should take into account time and be intuitive for clinicians. This paper uses Semantic Web technologies and HL7 Clinical Document Architecture (CDA) to provide well-defined interfaces that help clinicians to visualize the medical procedures performed and how clinical findings have changed over the time for a patient. To validate the proposal, the research has focused on diagnosis and clinical management of Glaucoma (Worldwide, it is the second leading cause of blindness) and the evaluation performed has involved health professionals who are not familiarized with Semantic Web technologies.

A collaborative workflow for building ontologies: A case study in the biomedical field

Much medical knowledge is contained within available literature, such as clinical guidelines and protocols. Recently, an interest has been developed in automatic content extraction to construct ontologies of this knowledge to make it more widely available. With groups of domain experts distributed geographically, and the growing amount of medical literature, an important challenge is to develop collaborative workflows to support ways for domain experts to contribute in the ontology learning process. This paper presents a collaborative workflow for ontology learning based on coupling an Ontology Learning Tool (OntoLancs) with and Ontology engineer (Protégé) to provide semi-automatic support for text mining and a collaborative tool to model formal ontologies. The work presented in this paper was evaluated with a case study on a Clinical Practice Guideline of Diabetic Retinopathy. The major benefits of coupling OntoLancs with Protégé are: a) a higher level of automation in the creation of domain ontologies and models, and b) strengthened communication and information exchange among domain experts that are physically distributed. Validations of user experiences indicate the applicability of our approach.

Enabling Reasoning on the Web: Introducing a Test-Bed Simulation Framework

The transformation of a document-based medical guideline into a computer-based decision support is a time-consuming and error-prone activity. One way to alleviate this burden is by automating as much as possible the knowledge intensive tasks within the life cycle of medical guidelines. This paper presents a test-bed simulation framework that aims to take advantages of both CommonKADS methodology and semantic Web technologies (OWL, SWRL, and OWL-S) to support automatic reasoning of knowledge-intensive tasks within guidelines. Based on the experiments conducted so far, the major benefit of the test-bed simulation is to enable experiments (simulations of clinical situations) that allow overcoming the main barriers to successfully express medical guidelines in an executable form compatible with electronic healthcare records.

Executing Medical Guidelines on the Web: Towards Next Generation Healthcare

There is still a lack of full integration between current Electronic Health Records (EHRs) and medical guidelines that encapsulate evidence-based medicine. Thus, general practitioners (GPs) and specialised physicians still have to read document-based medical guidelines and decide among various options for managing common non-life-threatening conditions where the selection of the most appropriate therapeutic option for each individual patient can be a difficult task. This paper presents a simulation framework and computational test-bed, called V.A.F. Framework, for supporting simulations of clinical situations that boosted the integration between Health Level Seven (HL7) and Semantic Web technologies (OWL, SWRL, and OWL-S) to achieve content layer interoperability between online clinical cases and medical guidelines, and therefore, it proves that higher integration between EHRs and evidence-based medicine can be accomplished which could lead to a next generation of healthcare systems that provide more support to physicians and increase patients’ safety.

Ontology-Based Information Extraction and Reservoir Computing for Topic Detection from Blogosphere’s Content: A Case Study About BBC Backstage

This research study aims at detecting topics and extracting themes (subtopics) from the blogosphere’s content while bridging the gap between the Social Web and the Semantic Web. The goal is to detect certain types of information from collections of blogs’ and microblogs’ narratives that lack explicit semantics. The approach presented introduces a novel approach that blends together two young paradigms: Ontology-Based Information Extraction (OBIE) and Reservoir Computing (RC). The novelty of the work lies in integrating ontologies and RC as well as the pioneering use of RC with social media data. Experiments with retrospect data from blogs and Twitter microblogs provide valuable insights into the BBC Backstage initiative and prove the viability of the approach presented in terms of scalability, computational complexity, and performance.

Extracting and Visualising Clinical Statements from Electronic Health Records

The use of Electronic Health Records (EHRs) standards, like the HL7 Clinical Document Architecture Release Two (CDA R2), presents challenges related to data management, information retrieval, and interactive visualisation. Indeed, it has been widely acknowledged that the usefulness of clinical data within EHRs is limited by the difficulty in accessing it. This paper demonstrates how key components of the Semantic Web (OWL and SPARQL) can tackle the problem of extracting patients’ clinical statements from EHRs. The novelty of the work resides in the exploitation of the terminology binding process, which specifies how to establish connections between elements of a specific terminology and an information model. The paper argues that in order to find new or better ways to visualise patients’ time-oriented information from EHRs, it is essential to perform first adequate EHR-specific searches. Towards this aim, the research presented proposes using SPARQL queries against formal semantic representations of clinical statements within HL7 CDA R2 documents in OWL. To validate the proposal, the study has focused on 1433 clinical statements that are within the physical examination section of 103 anonymised consultation notes in HL7 CDA R2. The paper adopts existing lightweight interactive visualisation techniques to represent patients’ clinical information in multiple consultations.

A case study on sepsis using PubMed and Deep Learning for ontology learning

We investigate the application of distributional semantics models for facilitating unsupervised extraction of biomedical terms from unannotated corpora. Term extraction is used as the first step of an ontology learning process that aims to (semi-)automatic annotation of biomedical concepts and relations from more than 300K PubMed titles and abstracts. We experimented with both traditional distributional semantics methods such as Latent Semantic Analysis (LSA) and Latent Dirichlet Allocation (LDA) as well as the neural language models CBOW and Skip-gram from Deep Learning. The evaluation conducted concentrates on sepsis, a major life-threatening condition, and shows that Deep Learning models outperform LSA and LDA with much higher precision.

Combining Semantic Web Technologies with Evolving Fuzzy Classifier eClass for EHR-Based Phenotyping: A Feasibility Study

In parallel to nation-wide efforts for setting up shared electronic health records (EHRs) across healthcare settings, several large-scale national and international projects are developing, validating, and deploying electronic EHR-oriented phenotype algorithms that aim at large-scale use of EHRs data for genomic studies. A current bottleneck in using EHRs data for obtaining computable phenotypes is to transform the raw EHR data into clinically relevant features. The research study presented here proposes a novel combination of Semantic Web technologies with the on-line evolving fuzzy classifier eClass to obtain and validate EHR-driven computable phenotypes derived from 1,956 clinical statements from EHRs. The evaluation performed with clinicians demonstrates the feasibility and practical acceptability of the approach proposed.

Incorporating Semantics into Data Driven Workflows for Content Based Analysis

Finding meaningful associations between text elements and knowledge structures within clinical narratives in a highly verbal domain, such as psychiatry, is a challenging goal. The research presented here uses a small corpus of case histories and brings into play pre-existing knowledge, and therefore, complements other approaches that use large corpus (millions of words) and no pre-existing knowledge. The paper describes a variety of experiments for content-based analysis: Linguistic Analysis using NLP-oriented approaches, Sentiment Analysis, and Semantically Meaningful Analysis. Although it is not standard practice, the paper advocates providing automatic support to annotate the functionality as well as the data for each experiment by performing semantic annotation that uses OWL and OWL-S. Lessons learnt can be transmitted to legacy clinical databases facing the conversion of clinical narratives according to prominent Electronic Health Records standards.

Enabling smooth integration between HL7 CDA and medical guidelines on the web with an initial focus on women’s health in general practice

Since the 1990s many researchers have proposed frameworks for modelling clinical guidelines and protocols in a computer-interpretable and computer-executable format. Nowadays, the various guideline representation languages and related frameworks also need to address compatibility with healthcare information systems that aim to be interoperable on nation-wide and even international-levels. This paper presents a simulation framework and computational test-bed, called V.A.F. Framework, that makes use of health level seven (HL7) and takes advantages of both CommonKADS methodology and Semantic Web technologies (OWL, SWRL, and OWL-S) to express a wide-spectrum of medical guidelines in an executable form compatible with HL7 clinical document architecture (CDA). To validate the proposal, which goes beyond interoperability of eHealth services and explores how to open up new capabilities in areas such as decision support and patient safety alerts, the research has focused on medical guidelines related to womenpsilas health in general practice.