Method for Accurate Unsupervised Cell Nucleus Segmentation
Abstract
To achieve the extreme accuracy rates demanded by applications in unsupervised automated cytology, it is frequently necessary to supplement the primary segmentation algorithm with a segmentation quality control system. The more robust the segmentation strategy, the less severe the data pruning need be at the segmentation validation stage. These issues are addressed as we describe our cell nucleus segmentation strategy which is able to achieve 100% accurate segmentation from a data set of 19946 cell nucleus images by automatically discarding the most difficult cell images. The automatic quality checking is applied to enhance the performance of a robust energy minimisation based segmentation scheme which already achieved a 99.47% accurate segmentation rate.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 25, 2001
- Accession Number
- ADA412347
Entities
People
- Brian Lovell
- Pascal Bamford
Organizations
- University of Queensland