Extensions to the time-oriented database model to support temporal reasoning in medical expert systems

Area of Science:

• Medical Informatics
• Computer Science
• Artificial Intelligence

Background:

• Physicians require temporal reasoning for clinical decisions.
• Current medical databases, like the Time-Oriented Databank (TOD) model, have limitations in storing and retrieving complex, time-sensitive patient data.
• Existing systems struggle to index patient data by multiple concurrent clinical events, hindering accurate analysis.

Purpose of the Study:

• To describe logical extensions to TOD-based databases for improved temporal reasoning in medical expert systems.
• To address limitations in storing and retrieving complex temporal information in medical database systems.
• To enhance the ability of medical expert systems to utilize time-varying clinical data.

Main Methods:

• Developed two object-oriented database extensions: temporal network (TNET) and extended temporal network (ETNET).
• Partitioned stored data into clinically relevant groupings (e.g., clinical visits, exacerbations).
• TNET associates observations with time intervals of clinical interest; ETNET adds executable reasoning methods triggered by event onset/conclusion.

Main Results:

• TNET and ETNET capture temporal relationships not represented in standard TOD databases.
• These extensions enable data retrieval based on multiple clinical contexts.
• ETNET allows dynamic modification of expert system reasoning based on clinical event timelines.

Conclusions:

• TNET and ETNET offer significant improvements for encoding and retrieving complex temporal relationships in patient data.
• These models enhance the temporal reasoning capabilities of medical expert systems.
• The developed systems facilitate more sophisticated medical decision-making by better representing clinical data over time.