Related Experiment Videos
Evaluation of an expert system producing geometric solids as output
C H Ketting1, M M Austin-Seymour, I J Kalet
1Radiation Oncology Dept., University of Washington, Seattle 98195, USA.
Proceedings. Symposium on Computer Applications in Medical Care
|January 1, 1995
Summary
Evaluating complex expert system outputs, like 3D geometric solids, presents challenges in setting standards and identifying deviations. Such systems may show lower performance scores and varied physician acceptance due to multiple factors.
Area of Science:
- Medical Informatics
- Computer Science
- Geometric Modeling
Background:
- Expert systems are increasingly used in medicine.
- Evaluating complex outputs, such as 3D geometric solids, poses unique challenges.
- Standardization and deviation classification are critical for assessing system performance.
Purpose of the Study:
- To evaluate an expert system generating three-dimensional geometric solids.
- To address the challenges in establishing comparison standards and classifying deviations for complex outputs.
- To assess the applicability of the evaluation design to other complex expert systems.
Main Methods:
- Utilized a panel of physicians to establish a comparison standard.
- Employed a separate panel of expert judges to identify and classify deviations.
- Focused on evaluating multi-parameter/multi-dimensional expert system outputs.
Main Results:
- Multi-dimensional expert system outputs can lead to lower overall performance scores.
- Increased variation in physician acceptability was observed with complex outputs.
- The observed effects were similar for both computer-generated and human-generated outputs.
Conclusions:
- The evaluation design is effective for complex expert system outputs.
- The complexity of expert system outputs influences performance scores and user acceptance.
- The methodology is applicable to a range of expert systems producing intricate outputs.