Comparative Study on Automated Cell Nuclei Segmentation Methods for Cytology Pleural Effusion Images

Area of Science:

• Medical Imaging
• Computational Pathology
• Biomedical Engineering

Background:

• Automated cell nuclei segmentation is vital for computer-aided diagnosis of cancer cells.
• Limited studies exist on automated analysis of pleural effusion cytology images due to a lack of reliable segmentation methods.

Purpose of the Study:

• To present a comparative study of twelve nuclei segmentation methods for cytology pleural effusion images.
• To evaluate the performance and efficiency of different segmentation techniques in this specific domain.

Main Methods:

• A comparative evaluation of twelve nuclei segmentation methods was performed.
• Each method included preprocessing, segmentation, and postprocessing steps.
• Quantitative evaluation used five performance metrics on 35 cytology images.

Main Results:

• Segmentation performances for Otsu, k-means, mean shift, Chan-Vese, and graph cut were 94%, 94%, 95%, 94%, and 93%, respectively.
• These methods demonstrated high abnormal nuclei detection rates.
• Average computational times per image varied significantly, with Otsu being the fastest at 1.08 seconds.

Conclusions:

• The study provides a quantitative comparison of nuclei segmentation methods for pleural effusion cytology.
• Findings are valuable for advancing automated analysis and computer-aided diagnosis in cytology.
• Selected methods offer promising results for future research in this area.