Example PhD

Temporal information extraction from medical records.

Supervisor: Dr I. Spasić

Keywords: Natural language processing, Temporal information extraction, Health informatics.

Reductionism pervades the medical sciences and the way diseases are diagnosed, treated and prevented. Time, space and context are vital for a system-level understanding, which through the use of computational tools may help develop modern treatments, which should ideally be individualized, minimized, multidimensional, time-sensitive, space-sensitive, synergistic and temporary. In this view, temporal information contained in medical records describes important aspects of medical treatments that need to be taken into account in medical problem-solving and decision-making. However, textual representation of such information is not structured in a machine readable format directly usable by data mining algorithms. Little work has been done in health informatics on temporal representation and reasoning. The aim of the current project would be to extract information that indicates whether a medication is started, stopped, or continued, when these events take place, and whether they are factual, suggested or conditional. Subsequently, the extracted information would be used for temporal ordering of the events described and placing them onto a time line in order to build a temporal representation of a medical treatment.

Key Skills/Background: Open to Computing Graduates and postgraduates.

Contact: Dr I. Spasić to discuss this research topic.