A High-Efficiency and High-Accuracy Cellular Segmentation Scheme for Imperfect Cytoarchitecture Images

Area of Science:

• Biomedical Image Analysis
• Computational Biology
• Cellular Imaging

Background:

• Accurate cellular segmentation is crucial for cell morphology analysis and disease diagnosis.
• Manual segmentation is error-prone, and general deep learning models struggle with imperfect cell images.
• There is a need for efficient and accurate cellular segmentation methods for challenging image datasets.

Purpose of the Study:

• To develop a high-efficiency and high-accuracy cellular segmentation scheme for imperfect cytoarchitecture images.
• To improve the reliability of cell morphology analysis and disease diagnosis through advanced segmentation techniques.
• To provide a robust solution for quantitative analysis in biomedical research.

Main Methods:

• Image enhancement techniques were applied to improve cell image quality.
• The Cellpose algorithm, utilizing the Cyto3 pretrained weight module, was employed as the core segmentation model.
• The proposed scheme requires no additional training, ensuring operational efficiency.

Main Results:

• The cellular segmentation scheme achieved high accuracy, with an IoU index of 0.86, ACC of 0.98, MCC of 0.91, and Dice of 0.93.
• The method successfully demonstrated quantitative differences in cell distribution density across mouse brain regions.
• Significant improvements in segmentation accuracy were observed compared to general algorithms.

Conclusions:

• The proposed cellular segmentation scheme offers high efficiency and accuracy for imperfect cell images.
• This method has significant potential for accurate biomedical research, including cell distribution density analysis and cellular localization.
• The approach provides a valuable tool for quantitative analysis in neuroscience and other cell-based research fields.