Development of Decision Support Algorithms for Intensive Care Medicine: A New Approach Combining Time Series Analysis and a Knowledge Base System with Learning and Revision Capabilities (original) (raw)
Related papers
Proceedings of the Amia Symposium, 2001
The overwhelming flood of data in intensive care medicine precludes consistent judgement of medical interventions by humans. Therefore, computerized decision support is needed to assist the health care professional in making reproducible, high-quality decisions at the bedside. Traditional expert systems rely on a tedious, labor-intensive and time-consuming approach in their development which falls short of exploiting existing numerical and qualitative data in large medical databases. Therefore, we applied a new concept of combining time series analysis and a knowledge base system with learning and revision capabilities (MOBAL) for rapid development of decision support algorithms for hemodynamic management of the critically ill. This approach could be successfully implemented in an existing intensive care database handling time-oriented data to validate and refine the intervention rules. The generation of hypotheses for identified contradictions lead to conclusive medical explanations that helped to further refine the knowledge base. This approach will provide for a more efficient and timely development of decision support algorithms.
Closed loop knowledge discovery for decision support in intensive care medicine
Clinical Decision Support Systems (CDSS) are becoming commonplace. They are used to alert doctors about drug interactions, to suggest possible diagnostics and in several other clinical situations. One of the approaches to building CDSS is by using techniques from the Knowledge Discovery from Databases (KDD) area. However using KDD for the construction of the knowledge base used in such systems, while reducing the maintenance work still demands repeated human intervention. In this work we present a KDD based architecture for CDSS for intensive care medicine. By resorting to automated data acquisition our architecture allows for the evaluation of the predictions made and subsequent action aiming at improving the predictive performance thus closing the KDD loop.
Adaptive knowledge discovery for decision support in intensive care units
Clinical Decision Support Systems (CDSS) are becoming commonplace. They are used to alert doctors about drug interactions, to suggest possible diagnostics and in several other clinical situations. One of the approaches to building CDSS is by using techniques from the Knowledge Discovery from Databases (KDD) area. However using KDD for the construction of the knowledge base used in such systems, while reducing the maintenance work still demands repeated human intervention. In this work we present a KDD based architecture for CDSS for intensive care medicine. By resorting to automated data acquisition our architecture allows for the evaluation of the predictions made and subsequent action aiming at improving the predictive performance thus enhancing adaptive capacities.
A New Real Time Clinical Decision Support System Using Machine Learning for Critical Care Unit
In intensive care units, mean arterial pressure (MAP) is a key clinical measure for estimating the health of critically ill patients. As a result, real-time clinical decisions support systems that detect irregularities and diversions in Chart can assist prevent major consequences by intervening early. At the bedside, state-of-the-art decision support systems are built on a three-phase system that includes offline training, transfer literacy, and retraining. With imprisonment and journey, their relationship in critical care units is difficult. Using a new machine literacy structure, we offer a real-time clinical decision assistance system vaticinating the Chart status at the bedside in this composition. The suggested system operates in real time at the bedside, eliminating the need for an offline training phase with big datasets. There are two levels to the proposed machine literacy framework. Stage I uses hierarchical temporal memory (HTM) and online literacy to enable real-time data processing and unsupervised prognostications. This is the first time it's been used on medical signals, to the best of our knowledge. Stage II is a long short-term memory (LSTM) classifier that predicts the status of the case's Chart in advance based on Stage I stream forecasts. We assess the proposed system's performance thoroughly and compare it to state-of-the-art systems that use logistic retrogression (LR). The proposed system beats LR in terms of bracket delicacy, recall, perfection, and area under the receiver operation wind, according to the comparison (AUROC).
Real-time Intelligent decision support in intensive medicine
2010
Daily, a great amount of data that is gathered in intensive care units, which makes intensive medicine a very attractive field for applying knowledge discovery in databases. Previously unknown knowledge can be extracted from that data in order to create prediction and decision models. The challenge is to perform those tasks in real-time, in order to assist the doctors in the decision making process. The Data Mining models should be continuously assessed and optimized, if necessary, to maintain a certain accuracy. In this paper we present the INTCare system, an intelligent decision support system for intensive medicine and the way it was adapted to the new requirements. Some preliminary results are analysed and discussed.
Intelligent decision support in Intensive Care Medicine
2006
This paper introduces the INTCare system, an intelligent decision support system for intensive medicine. The system aims at the automation of the Knowledge Discovery Process by using autonomous agents that are responsible for the various constituent steps. The system enables automation of data acquisition and model updating avoiding human intervention. We present the first impressions after the deployment of INTCare in a real environment (Intensive Care Unit of the Hospital de Santo António, Oporto, Portugal) where it is supporting the physicians' decisions by means of prognostic Data Mining models. In particular, these techniques are used to predict organ failure and mortality assessment. The main intention is to change the current reactive behaviour to a pro-active one, enhancing the Quality of Service.
1990
Our initial goal was to develop protocols to ensure uniform care in both groups of our current randomized clinical trial, "Extracorporeal C02 Removal for ARDS" (ECCO2R, NIH Grant HL36787). These hypoxemia protocols utilize the bedside intensive care unit (ICU) computer terminal to prompt the clinical care team with therapeutic and diagnostic suggestions. The protocols have been used for over 15,000 hours in 61 adult respiratory distress syndrome (ARDS) patients. 38 of these ARDS patients met extra corporeal membrane oxygenation (ECMO) criteria. The survival of the ECMO criteria ARDS patients was 41%,four times that expected (9%) from historical data (p < 0.003). The four fold increase in survival is surprising. The success of these computer protocols clearly established the feasib of controlling the therapy ofseverely ill patients. Over the lastfour years we have refined the process which we usefor generating computerized protocols. The purpose of this paper is to present the six step development strategy which we are successfully using to produce computerized critical care protocols.
1999
The paper describes a case study in combining di erent methods for acquiring medical knowledge. Given a huge amount of noisy, high dimensional numerical time series data describing patients in intensive care, the support vector machine is used to learn when and how to change the dose of which drug. Given medical knowledge about and expertise in clinical decision making, a rst-order logic knowledge base about e ects of therapeutical interventions has been built. As a preprocessing mechanism it uses another statistical method. The integration of numerical and knowledge-based procedures eases the task of validation in two w ays. On one hand, the knowledge base is validated with respect to past patients' records. On the other hand, medical interventions that are recommended by learning results are justi ed by the knowledge base.
Knowledge discovery and knowledge validation in intensive care
Artificial Intelligence in Medicine, 2000
Operational protocols are a valuable means for quality control. However, developing operational protocols is a highly complex and costly task. We present an integrated approach involving both intelligent data analysis and knowledge acquisition from experts that support the development of operational protocols. The aim is to ensure high quality standards for the protocol through empirical validation during the development, as well as lower development cost through the use of machine learning and statistical techniques. We demonstrate our approach of integrating expert knowledge with data driven techniques based on our effort to develop an operational protocol for the hemodynamic system.
2012
Patients in Cardiothoracic Intensive Care Units (CICU) are physiologically weak and require watchful monitoring and support. Such monitoring generates massive amount of data that enable early detection of changes in the patient's condition and provide information that help medical staff to give the treatment and evaluate the response to medical interventions. The countless data gathered from monitoring systems and clinical information systems have created a challenge and are time consuming for clinicians to analyze. This paper discusses the implementation of an intelligent system that has been designed to improve interpretation of clinical data which will then increase the quality and efficiency of the working environment in CICU. The implementation is based on the description state from the cardiologist. This research work extends the cardiologist approach by providing the heuristic rules-based approach to address the treatment. The system is intended to help physicians and CICU staffs to diagnose and track the conditions of patients.