MANAGING TEMPORAL WORLDS FOR MEDICAL TREND DIAGNOSIS

The medical trend diagnosis system TrenDx has been applied as a protot ype for diagnosing pediatric growth disorders, and as a proof of conce pt in detecting clinically significant trends in hemodynamics and bloo d gases in intensive care unit patients, TrenDx diagnoses trends by ma tching patient data to patterns of normal and abnormal trends called t rend templates that define disorders as typical patterns of relevant v ariables. These patterns consist of a partially ordered set of tempora l intervals with uncertain endpoints. Bound to each temporal interval are value constraints on real-valued functions of measurable parameter s, The temporal uncertainty in trend templates allows TrenDx to conclu de both what trend pattern best matches the data and also when signifi cant landmarks and phase transitions have occurred within the best mat ching 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 nu mber 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 differen t points in the trade-off between accuracy of trend detection and algo rithm efficiency.