Related Experiment Videos
Bayesian belief networks in quantitative histopathology.
P H Bartels1, D Thompson, M Bibbo
1Optical Sciences Center, University of Arizona, Tucson 85721.
Analytical and Quantitative Cytology and Histology
|December 1, 1992
Summary
Bayesian belief networks offer dynamic range and numeric responses ideal for classification. These networks enable accurate classification decisions even with overlapping feature data, demonstrating their utility in automated histopathology.
Area of Science:
- Computational Biology
- Digital Pathology
- Machine Learning
Background:
- Descriptive classification schemes require methods with dynamic range and precise numeric responses.
- Traditional classification methods may struggle with features exhibiting overlapping tolerance regions.
Purpose of the Study:
- To analyze the numeric response characteristics of Bayesian belief networks.
- To demonstrate the application of Bayesian belief networks as control modules for automated scene segmentation in histopathology.
Main Methods:
- Analysis of Bayesian belief network numeric response characteristics.
- Implementation of Bayesian belief networks for automated scene segmentation in histopathology.
Main Results:
- Bayesian belief networks possess suitable dynamic range and numeric response characteristics for classification.
- Cumulative use of features with overlapping tolerance regions can lead to unequivocal classification decisions.
- Demonstrated successful application as control modules in automated histopathology scene segmentation.
Conclusions:
- Bayesian belief networks are well-suited for descriptive classification tasks.
- Their numeric properties facilitate robust classification, even with ambiguous feature data.
- These networks represent a powerful tool for advancing automated histopathology image analysis.