Ontology acquisition from on-line knowledge sources.
(57/4007)
Electronic knowledge representation is becoming more and more pervasive both in the form of formal ontologies and less formal reference vocabularies, such as UMLS. The developers of clinical knowledge bases need to reuse these resources. Such reuse requires a new generation of tools for ontology development and management. Medical experts with little or no computer science experience need tools that will enable them to develop knowledge bases and provide capabilities for directly importing knowledge not only from formal knowledge bases but also from reference terminologies. The portions of knowledge bases that are imported from disparate resources then need to be merged or aligned to one another in order to link corresponding terms, to remove redundancies, to resolve logical conflicts. We discuss the requirements for ontology-management tools that will enable interoperability of disparate knowledge sources. Our group is developing a suite of tools for knowledge-base management based on the Protege-2000 environment for ontology development and knowledge acquisition. We describe one such tool in detail here: an application for incorporating information from remote knowledge sources such as UMLS into a Protege knowledge base. (+info)
Exploiting multi-modal reasoning for knowledge management and decision support: an evaluation study.
(58/4007)
We present the first evaluation results of a knowledge management and decision support system for Type I diabetes patients' care. Such system, meant to help physicians in therapy revision, relies on the integration of Rule Based Reasoning and Case Based Reasoning, and exploits both explicit and implicit knowledge. Reliability was positively judged by a group of expert diabetologists; an increase in its performances is foreseen as new knowledge will be acquired, through the system usage in clinical practice. (+info)
Design of a clinical alert system to facilitate development, testing, maintenance, and user-specific notification.
(59/4007)
Creation and maintenance of electronic clinical alerts within a hospital's electronic medical record (EMR) or database poses a number of challenges. Development can require significant programming effort. Final testing should ideally be performed in a real clinical environment without clinician notification, which may create technical challenges. After an alert is in production, modifications may become necessary in response clinician feedback, changes in clinical factors, or technical issues. Changes may be required in the knowledge base utilized by the alert or in the presentation of the alert condition to the clinicians. Occasionally, different users within the clinical environment may wish to have the same alert data presented differently. We have developed a strategy which allows development of multi-functional alerts and facilitates modification of alert function and/or presentation with minimal to no programming effort. Some elements of this scheme may be appropriate for incorporation into clinical alerting standards. (+info)
QueryCat: automatic categorization of MEDLINE queries.
(60/4007)
A searcher's inability to formulate an appropriate query can result in an overwhelming number of retrieved documents. Our approach to this problem is to use information about common types or categories of queries to (1) reformulate the user's initial query and (2) create an informative organization of the retrieved documents from the reformulated query. To achieve these goals, we first must identify which common categories or types of queries are the best abstraction of the user's specific query. In this paper, we describe a system that performs this first step of categorizing the user's query. Our system uses a two-phased approach: a lexical analysis phase, and a semantic analysis phase. An evaluation of our system demonstrates that its query categorization corresponds reasonably well to the query categorizations by medical librarians and physicians. (+info)
A knowledge-based patient assessment system: conceptual and technical design.
(61/4007)
This paper describes the design of an inpatient patient assessment application that captures nursing assessment data using a wireless laptop computer. The primary aim of this system is to capture structured information for facilitating decision support and quality monitoring. The system also aims to improve efficiency of recording patient assessments, reduce costs, and improve discharge planning and early identification of patient learning needs. Object-oriented methods were used to elicit functional requirements and to model the proposed system. A tools-based development approach is being used to facilitate rapid development and easy modification of assessment items and rules for decision support. Criteria for evaluation include perceived utility by clinician users, validity of decision support rules, time spent recording assessments, and perceived utility of aggregate reports for quality monitoring. (+info)
Knowledge representation and tool support for critiquing clinical trial protocols.
(62/4007)
The increasing complexities of clinical trials have led to increasing costs for investigators and organizations that author and administer those trials. The process of authoring a clinical trial protocol, the document that specifies the details of the study, is usually a manual task, and thus authors may introduce subtle errors in medical and procedural content. We have created a protocol inspection and critiquing tool (PICASSO) that evaluates the procedural aspects of a clinical trial protocol. To implement this tool, we developed a knowledge base for clinical trials that contains knowledge of the medical domain (diseases, drugs, lab tests, etc.) and of specific requirements for clinical trial protocols (eligibility criteria, patient treatments, and monitoring activities). We also developed a set of constraints, expressed in a formal language, that describe appropriate practices for authoring clinical trials. If a clinical trial designed with PICASSO violates any of these constraints, PICASSO generates a message to the user and a list of inconsistencies for each violated constraint. To test our methodology, we encoded portions of a hypothetical protocol and implemented designs consistent and inconsistent with known clinical trial practice. Our hope is that this methodology will be useful for standardizing new protocols and improving their quality. (+info)
Event discovery in medical time-series data.
(63/4007)
Vast amounts of clinical information are generated daily on patients in the health care setting. Increasingly, this information is collected and stored for its potential utility in advancing health care. Knowledge-based systems, for example, might be able to apply rules to the collected data to determine whether a patient has a certain condition. Often, however, the underlying knowledge needed to write such rules is not well understood. How could these clinical data be useful then? Use of machine learning is one answer. We present a pipeline for discovering the knowledge needed for event detection in medical time-series data. We demonstrate how this process can be applied in the development of intelligent patient monitoring for the intensive care unit (ICU). Specifically, we develop a system for detecting Otrue alarmO situations in the ICU, where currently as many as 86% of bedside monitor alarms are false. (+info)
Boosting naive Bayesian learning on a large subset of MEDLINE.
(64/4007)
We are concerned with the rating of new documents that appear in a large database (MEDLINE) and are candidates for inclusion in a small specialty database (REBASE). The requirement is to rank the new documents as nearly in order of decreasing potential to be added to the smaller database as possible, so as to improve the coverage of the smaller database without increasing the effort of those who manage this specialty database. To perform this ranking task we have considered several machine learning approaches based on the nai ve Bayesian algorithm. We find that adaptive boosting outperforms nai ve Bayes, but that a new form of boosting which we term staged Bayesian retrieval outperforms adaptive boosting. Staged Bayesian retrieval involves two stages of Bayesian retrieval and we further find that if the second stage is replaced by a support vector machine we again obtain a significant improvement over the strictly Bayesian approach. (+info)