Knowledge-Based Temporal Abstraction in Clinical-Management Tasks
Yuval Shaḥar, Amar K. Das, Samson W. Tu, Mark A. Musen · 1994
We describe a general method for abstracting higherlevel, interval-based concepts from time-stamped clinical data, the knowledge-based temporal-abstraction [KBTA] method. We focus on the knowledge representation, knowledge acquisition and knowledge reuse and sharing aspects of the KBTA method. We describe five mechanisms which solve the five tasks composing the KBTA method, and four types of knowledge necessary for instantiating these mechanisms in a particular domain. We present an example of instantiating the KBTA method in the domain of monitoring insulin-dependent diabetes patients. 1. The Temporal-Abstraction Task There are several desired features for a method solving the TA task. The input and output parameter values might be of different types (e.g., numbers, symbols) and abstraction levels (e.g., BLOOD GLUCOSE LEVEL 74; GLUCOSE STATE = LOW). Input data might arrive out of temporal order, and interpretations should be revised accordingly. Several alternate interpretations might need to be maintained and followed. From the knowledge-representation aspect, acquiring necessary knowledge from domain experts should be facilitated, as well as maintenance of that knowledge. Reusing the domain-independent abstraction knowledge for solving the TA task in other domains should be possible, as well as sharing some of the domain-specific knowledge with other tasks in the same domain.