Nucleus segmentation of cervical cytology images based on multi-scale fuzzy clustering algorithm

Area of Science:

• Medical Imaging
• Computational Pathology
• Biomedical Engineering

Background:

• Accurate nucleus segmentation in cervical cell images is crucial for early cervical cancer diagnosis.
• Challenges include overlapping cells, uneven staining, and poor contrast, hindering precise segmentation.

Purpose of the Study:

• To develop an advanced segmentation method for cervical nuclei.
• To enhance the accuracy and reliability of nucleus identification in cervical cancer screening.

Main Methods:

• A multi-scale fuzzy clustering algorithm is proposed for segmenting cervical cell clump images at various scales.
• A novel area-prior-based interesting degree is introduced to measure node significance.
• These methods address the challenge of determining the number of clusters and improve nucleus recognition.

Main Results:

• The proposed method effectively segments cervical nuclei, demonstrating superior performance compared to existing state-of-the-art algorithms.
• High accuracy nucleus segmentation results were achieved on the IBSI2014 and IBSI2015 public datasets.
• The algorithm successfully overcomes common segmentation difficulties like overlapping nuclei and poor image quality.

Conclusions:

• The developed multi-scale fuzzy clustering algorithm offers significant advantages for cervical nucleus segmentation.
• This method enhances the accuracy of early cervical cancer diagnosis through improved image analysis.
• The approach provides a robust solution for nucleus segmentation in challenging cervical cell images.