ions of time-stamped clinical data are useful for planning therapy, for monitoring therapy, and for creating high-level summaries of time-oriented clinical databases. Temporal abstractions also support explanations by an intelligent patient-record system and can be used for representation of the goals and intentions of clinical guidelines and protocols. We propose to reengineer and expand the scope of the RESUME system, a prototype computer program that implements the knowledge-based temporal- abstraction method, a conceptual and computational framework that we have developed for abstraction of time-stamped clinical data into clinically meaningful interval-based concepts. RESUME has been evaluated with highly encouraging results in several clinical areas. We will address the practical and theoretical issues of representation, acquisition, maintenance, and reuse of temporal-abstraction knowledge.
Our specific aims are defined by a four-step research plan: 1. We will define formally the knowledge requirements for five computational modules (mechanisms) we employ, thus facilitating the acquisition, maintenance, reuse, and sharing of the required knowledge. 2. We will enhance, expand, and redesign five computational temporal- abstraction mechanisms: (a) Automatic formation of meaningful contexts for interpretation of clinical data. (b) Classification of clinical data that have equivalent time stamps into higher-level concepts. (c) Temporal inference (e.g., the join of certain interval-based clinical abstractions into longer ones). (d) Interpolation between temporally disjoint clinical abstractions, including a development of a probabilistic representation and semantics. (e) Matching of predefined and runtime temporal patterns, given time- stamped data and conclusions. 3. We will develop a tool for automated acquisition, from expert physicians, of temporal-abstraction knowledge, using techniques from the PROTEGE-II project for designing knowledge-based systems. 4. We will validate and evaluate our methodology and its implementation. (a) We will assess the value of the knowledge-acquisition tool in several experiments. (b) We will validate the performance of the computational mechanisms in the domain of therapy of patients who have insulin-dependent diabetes by collaboration with expert endocrinologists. (c) We will evaluate the overall framework within EON, a project in which researchers are implementing an integrated architecture for protocol-based care.
Showing the most recent 10 out of 19 publications
