Extracting narratives from clinical patient data

Narratives can express: (a) human-understandable raw data narrations, suitable for explanation, decision and prediction tasks; (b) background knowledge and requirements.
Leveraging expertise gained in “Natural Language Processing” and “Process Mining”, we will define a methodology for constructing both narrative kinds to describe the history of one or multiple patients in terms of meaningful event types (concerning diagnoses, treatments, etc.).
Two narrative-extraction tasks will be tackled: (i) compute a “care-flow interpretation” (i.e. a sequence of event types’ instances) for one patient; and (ii) compute “care-flow patterns” for a patient group.
A-priori “narrative-like” domain knowledge (e.g., modelling physical/organizational constraints) will be possibly used in both tasks –so that humans and AI tools can jointly create/adapt the clinical narratives.
The discovered narratives are expected to facilitate the comprehension and analysis of care-flow processes, and possibly enable semantic-level reasoning/prediction tasks (e.g., forecast future events, or check compliance to guidelines).

More details available in https://drive.google.com/file/d/16U1OfH8LPBq5niuZj6lZ85j3OjDcv9kI/view?usp=sharing

Output

conference paper

Project Partners:

• Consiglio Nazionale delle Ricerche (CNR), Italy, luigi pontieri
• FBK, Chiara Ghidini
• INESC TEC, Alípio M. Jorge
• CNR, Luigi Pontieri

Primary Contact: luigi pontieri, Consiglio Nazionale delle Ricerche (CNR), Italy