Related Experiment Videos
Machine vision-based histometry of premalignant and malignant prostatic lesions
P H Bartels1, D Thompson, H G Bartels
1Optical Sciences Center, University of Arizona, Tucson, USA.
Pathology, Research and Practice
|September 1, 1995
Summary
Automated analysis of prostate cancer images is now possible using knowledge-guided control for processing and segmentation. This approach links histopathologic terms to computed measurements for improved diagnostic accuracy.
Area of Science:
- Digital pathology
- Computational imaging
- Medical artificial intelligence
Background:
- Histopathologic image analysis for prostate cancer diagnosis is complex.
- Automated interpretation requires linking visual features to diagnostic criteria.
Purpose of the Study:
- To enable automated analysis and interpretation of prostatic histopathologic images.
- To establish correspondence between histopathologic concepts and computed histometric entities.
Main Methods:
- Implementation of knowledge-guided control for image processing and segmentation.
- Introduction of "interpretive transforms" to link concepts and measurements.
- Utilizing an expert system with a model-based reasoning process for scene segmentation control.
- Structuring the expert system as an associative network with frames, controlling a knowledge file and image processing algorithms.
Main Results:
- Automated analysis and interpretation of prostatic histopathologic images are achievable.
- A framework for establishing correspondence between histopathologic terms and computed entities has been developed.
Conclusions:
- Knowledge-guided control facilitates automated histopathologic image analysis.
- The "interpretive transforms" and expert system approach enhance the link between pathology and computational analysis for prostate lesions.