| Literature DB >> 7799818 |
K Kuhn1, T Zemmler, M Reichert, D Rösner, O Baumiller, H Knapp.
Abstract
A routinely used system for report generation, based on direct physician data entry, has been combined with a knowledge-based module. The knowledge-based system is operating in the background, and guides the user by dynamically suggesting diagnoses and generating hints regarding the actual and possible further diagnostic procedures. Hybrid technology with neural networks and rules is used. A laboratory evaluation has shown good agreement between diagnoses suggested by the system and confirmed diagnoses, with a kappa coefficient of 0.85 and an area under the ROC curve of 0.92. In a survey following a 14-day test period, physicians stated that the system was both sensible and helpful. The current application domain is ultrasound reporting.Mesh:
Year: 1994 PMID: 7799818
Source DB: PubMed Journal: Methods Inf Med ISSN: 0026-1270 Impact factor: 2.176