Biomedical subjects

I J Haimowitz

Publications and source records attributed to I J Haimowitz.

6 recordsLinked to original sources

PubMed1996

Managing temporal worlds for medical trend diagnosis.

The medical trend diagnosis system TrenDx has been applied as a prototype for diagnosing pediatric growth disorders, and as a proof of concept in detecting clinically significant trends in hemodynamics and blood gases in intensive care unit patients. TrenDx diagnoses trends by matching patient data to patterns of normal and abnormal trends called trend templates that define disorders as typical patterns of relevant variables. These patterns consist of a partially ordered set of temporal intervals with uncertain endpoints. Bound to each temporal interval are value constraints on real-valued functions of measurable parameters. The temporal uncertainty in trend templates allows TrenDx to conclude both what trend pattern best matches the data and also when significant landmarks and phase transitions have occurred within the best matching trend. The temporal uncertainty in trend templates requires that TrenDx consider alternate temporal worlds in monitoring patient data. The number of temporal worlds grows worst case polynomially in the number of time slices of data. To manage the competing temporal worlds, TrenDx employs two techniques: beam search based on regression scores, and temporal granularity in the trend template definitions. These two techniques, described here in detail, allow TrenDx to choose different points in the trade-off between accuracy of trend detection and algorithm efficiency.

Algorithms↗

PubMed1995

Clinical monitoring using regression-based trend templates.

Our computer program TrenDx detects clinically significant trends in time-ordered patient data by matching data to patterns called trend templates, denoting multivariate temporal and value variation in normality and in disease. Previously a purely constraint-based TrenDx diagnosed pediatric growth trends and reached the same diagnoses as a panel of experts, at a time no later than the experts, in most of 30 cases. Improvement required resolving outstanding representational issues. In this paper we describe regression-based trend templates, updated TrenDx algorithms, and their application to monitoring intensive care unit and pediatric growth data. We focus on new results in diagnosing pediatric growth trends, and discuss potential application domains for TrenDx.

Algorithms↗

PubMed1994

Intelligent diagnostic monitoring using trend templates.

In previous work we have defined our trend template epistemology for clinically significant trends and we have illustrated and tested a program TrenDx that monitors time-ordered process data by matching the data to trend templates. Our initial application domain was pediatric growth monitoring. In continuing work we have explored monitoring hemodynamic and respiratory parameters of intensive care unit patients. This application has highlighted the needs for advances in our representation and monitoring algorithms. In particular, we have added reasoning with uncertainty to the trend template epistemology, and a new control structure allowing numerical ranking of competing trend templates. Furthermore, intelligent monitoring in any medical domain requires a coherent framework for diagnostic monitoring. In this paper we show how TrenDx can be extended to a framework including sending alarms, changing clinical context, and filtering data streams.

Diagnosis, Computer-Assisted↗

PubMed1993

Hypothesis-driven data abstraction with trend templates.

We have written a prototype computer program called TrenDx for automated trend detection during process monitoring. The program uses a representation called trend templates that define disorders as typical patterns of relevant variables. These patterns guide the assignment of primary data to abstracted intervals or phases of the monitored process. TrenDx has been applied to the task of pediatric growth monitoring. The results of an exploratory clinical trial are reviewed here. The general utility of TrenDx for clinical trend detection and diagnosis is illustrated with examples from several task domains.

Child↗

PubMed1991

Modeling all dialogue system participants to generate empathetic responses.

A dialogue system between an expert system and its users is described which combines two recent hypotheses. First, that the dialogue system should explicitly model both the person directly interacting with the dialogue system (the agent) and the person reasoned about by the expert system (the patient) in order to communicate meaningfully with both people. Second, that a dialogue system can model the domain-related beliefs, preferences and concerns of both its users and generate responses empathetic to both. This dialogue system is called SERUM, standing for 'System for Empathetic Responses with User Models.' SERUM generates natural-language responses about attribute values of domain objects, via three transformations. First, the system converts properties of the agent and patient, and domain knowledge, into a pragmatic objective like empathy. Second, SERUM converts the pragmatic objectives into surface structure cues like object emphasis and level of technicality. Finally, SERUM converts the surface structure cues to realize text that is natural, appropriately technical and emotionally empathetic. SERUM is demonstrated in describing tests and treatments for lung disease in AIDS patients, a sensitive domain where empathetic responses may be needed.

Algorithms↗

PubMed1991

Influences on the performance of hospital clinical event monitoring.

The implementation of real-time clinical monitors in large hospital information systems places large performance demands on these systems. Meeting these demands not only requires methodologies to augment the performance of individual monitors but an understanding of how patient population and monitor characteristics might influence overall system performance. We have built a multimonitor simulator to study these influences on performance. In doing so, we have focused on the impact of a variety of techniques to cache a subset of a large number of monitors in primary memory.

Boston↗