Comparing expert systems for identifying chest x-ray reports that support pneumonia.
(17/4007)
We compare the performance of four computerized methods in identifying chest x-ray reports that support acute bacterial pneumonia. Two of the computerized techniques are constructed from expert knowledge, and two learn rules and structure from data. The two machine learning systems perform as well as the expert constructed systems. All of the computerized techniques perform better than a baseline keyword search and a lay person, and perform as well as a physician. We conclude that machine learning can be used to identify chest x-ray reports that support pneumonia. (+info)
Evaluating variable selection methods for diagnosis of myocardial infarction.
(18/4007)
This paper evaluates the variable selection performed by several machine-learning techniques on a myocardial infarction data set. The focus of this work is to determine which of 43 input variables are considered relevant for prediction of myocardial infarction. The algorithms investigated were logistic regression (with stepwise, forward, and backward selection), backpropagation for multilayer perceptrons (input relevance determination), Bayesian neural networks (automatic relevance determination), and rough sets. An independent method (self-organizing maps) was then used to evaluate and visualize the different subsets of predictor variables. Results show good agreement on some predictors, but also variability among different methods; only one variable was selected by all models. (+info)
Workflow analysis and evidence-based medicine: towards integration of knowledge-based functions in hospital information systems.
(19/4007)
The large extent and complexity of scientific evidence described in the concept of evidence-based medicine often overwhelms clinicians who want to apply best external evidence. Hospital Information Systems usually do not provide knowledge-based functions to support context-sensitive linking to external information sources. Knowledge-based components need specific data, which must be entered manually and should be well adapted to clinical environment to be accepted by clinicians. This paper describes a workflow-based approach to understand and visualize clinical reality as a preliminary to designing software applications, and possible starting points for further software development. (+info)
Users' evaluation of OncoDoc, a breast cancer therapeutic guideline delivered at the point of care.
(20/4007)
Despite the dissemination of computer-based "clinical practice guidelines" as decision support systems, low practical compliance rates are still observed. The reason commonly invoked is that such recommendations, suited to average patients, are not rules for all the patients. Rather than providing automatic decision support, OncoDoc allows the clinician to operationalize the implemented breast cancer therapeutic expertise through his hypertextual reading of the knowledge base. In this way, he has the opportunity to interpret the information provided in the context of his patient therefore controlling his categorization to the closest appropriate "average patient". After a four-month real-life experimentation of the system, a survey was conducted among the users. The observed compliance, significantly higher than the best figures found in the literature, and the clinicians objective and subjective evaluation of the system reinforced the implementation choices adopted in OncoDoc. (+info)
Justification of automated decision-making: medical explanations as medical arguments.
(21/4007)
People use arguments to justify their claims. Computer systems use explanations to justify their conclusions. We are developing WOZ, an explanation framework that justifies the conclusions of a clinical decision-support system. WOZ's central component is the explanation strategy that decides what information justifies a claim. The strategy uses Toulmin's argument structure to define pieces of information and to orchestrate their presentation. WOZ uses explicit models that abstract the core aspects of the framework such as the explanation strategy. In this paper, we present the use of arguments, the modeling of explanations, and the explanation process used in WOZ. WOZ exploits the wealth of naturally occurring arguments, and thus can generate convincing medical explanations. (+info)
Integrated knowledge-based functions in a hospital cancer registry--specific requirements for routine applicability.
(22/4007)
The background of the presented work is the design, realization, and routine use of integrated knowledge-based functions in the context of a hospital cancer registry. The first field of application was supporting registrars to detect data inconsistencies and incompleteness timely during the documentation process. Especially, we focused on the acceptance of the administrator of the underlying information system and on the phenomenon of duplicate and outdated messages. These aspects are specific for integrated knowledge based functions and a precondition for obtaining a routine applicability and acceptance. (+info)
Intelligent split menus for data entry: a simulation study in general practice medicine.
(23/4007)
A compelling notion in menu design is that a few of the most frequently selected items should be placed as a hot list at the top of the menu. A few researchers have explored this type of interface control, known as a split menu, and have investigated the identification of the hot-list items by statistical analysis of past data. We extend the technique to automated development of dynamic hot-lists for entry of medication data in a General Practice setting. Using clinical data from 113,000 visits, a statistical model is developed and evaluated by simulated data entry of cases held back from training. Simulated SOAP note entry shows 12-item hot lists to hold over 70% of desired drug and diagnosis selections. Intelligent split menus should improve user efficiency if current selection methods require 3 seconds or more per item. A demonstration prototype can be downloaded over the Web. (+info)
Classification algorithms applied to narrative reports.
(24/4007)
Narrative text reports represent a significant source of clinical data. However, the information stored in these reports is inaccessible to many automated decision support systems. Data mining techniques can assist in extracting information from narrative data. Multiple classification methods, such as rule generation, decision trees, Bayesian classifiers, and information retrieval were used to classify a set of 200 chest X-ray reports according to 6 clinical conditions indicated. A general-purpose natural language processor was used to convert the narrative text into a coded form that could be used by the classification algorithms. Significant differences in performance were found between algorithms. The best performing algorithm applied to the processor output was significantly better than information retrieval applied to raw text. Predictor variables from the coded processor output were limited to avoid overfitting. Methods that limited by domain knowledge performed significantly better than those that limited by conditional probabilities of the variables in the training set. Algorithms were also shown to be dependent on training set size. (+info)