Logic-Based Event Detection from Timestamped Medical Data
A novel framework grounded in logic identifies complex, time-extended events from timestamped information and contextual knowledge. By applying logical rules, it defines the initiation and conclusion criteria for basic temporal events, which are then integrated into meta-events. In the medical field, episodes of illness and treatments are derived from clinical data such as diagnoses and medication records, subsequently merged into more comprehensive disease events. A correction mechanism identifies the most consistent sets of events when erroneous ones are detected. Although complete reasoning is impractical, specific constraints allow for polynomial-time data complexity. A prototype system has been developed to implement essential components utilizing answer set programming.
Key facts
- Novel logic-based approach to detecting high-level temporally extended events
- Uses logical rules for existence and termination conditions
- Applied to medical domain: disease episodes and therapies from clinical data
- Constraints identify incompatible event combinations
- Repair mechanism selects preferred consistent event sets
- Full reasoning is intractable
- Certain restrictions ensure polynomial-time data complexity
- Prototype system uses answer set programming
Entities
—