Automated selection of clinical data to support radiographic interpretation

Area of Science:

• Medical Informatics
• Radiology
• Artificial Intelligence in Medicine

Background:

• Computerized hospital information systems (HIS) accumulate vast amounts of clinical data.
• Efficient retrieval and synthesis of this data for clinical tasks, such as radiographic interpretation, can be challenging.
• Information overload can hinder the effective use of HIS data.

Purpose of the Study:

• To develop an automated process for abstracting and summarizing salient clinical data.
• To create a concise report to aid radiologists in interpreting chest x-rays.
• To address the inefficiency of using large volumes of data in HIS.

Main Methods:

• Utilizing a combination of expert systems technology and information theory for data selection.
• Developing a reporting system to abstract select data from a clinical database.
• Implementing pre-compiled tables of finding-specific information contents to optimize processing time.

Main Results:

• The system abstracts relevant clinical data to produce a brief summary report.
• Initial versions were time-prohibitive due to run-time expert systems usage.
• Optimization using pre-compiled tables significantly reduced processing time, making the application clinically feasible.

Conclusions:

• Automated clinical data abstraction and summarization can enhance radiographic interpretation.
• Expert systems and information theory are effective tools for selective data retrieval.
• Optimized processing times are crucial for the clinical feasibility of such information systems.