Segmentation of cervical cell nuclei in high-resolution microscopic images: A new algorithm and a web-based software
Christoph Bergmeir1, Miguel García Silvente, José Manuel Benítez
1Department of Computer Science and Artificial Intelligence, E.T.S. de Ingenierías Informática y de Telecomunicación, University of Granada, Granada, Spain. c.bergmeir@decsai.ugr.es
Computer Methods and Programs in Biomedicine
|February 7, 2012
Summary
Automating cervical cancer screening requires accurate cell nuclei segmentation. This study introduces a novel algorithm for precise nuclei segmentation in high-resolution microscopy images, improving diagnostic accuracy.
Area of Science:
- Digital pathology
- Biomedical image analysis
- Computational biology
Background:
- Automating cervical cancer screening is crucial for public health.
- Accurate segmentation of cell nuclei in microscopy images presents a significant challenge, especially in high-resolution scans with varying conditions.
- Existing methods struggle with segmenting numerous nuclei with diverse characteristics.
Purpose of the Study:
- To develop and implement an automated and interactive system for segmenting cell nuclei in high-resolution cervical cancer screening images.
- To propose a novel algorithm for robust and accurate nuclei segmentation, adaptable to various image characteristics and user control.
- To facilitate data storage and expert interaction through a web-based architecture.
Main Methods:
- A novel algorithm combining a voting scheme with prior knowledge for nuclei localization.
- Elastic segmentation and level set algorithms for precise nuclei shape determination.
- Image processing techniques including mean-shift filtering, median filtering, Canny edge detection, and randomized Hough transform for ellipse detection.
Main Results:
- The implemented system processes full-resolution images and offers both automatic and interactive segmentation modes.
- The proposed algorithm demonstrated promising results in segmenting cell nuclei across a diverse dataset of 207 images.
- The system supports data storage and collaboration between technical and medical experts via a web-based platform.
Conclusions:
- The developed system and algorithm offer a robust solution for automated cell nuclei segmentation in cervical cancer screening.
- The approach enhances the accuracy and efficiency of analyzing high-resolution microscopy slides.
- The web-based architecture promotes interdisciplinary collaboration, advancing digital pathology applications.
