Cell segmentation in fluorescence microscopy images based on multi-scale histogram thresholding.

Area of Science:

• Biomedical Imaging
• Computational Biology
• Image Analysis

Background:

• Accurate cell segmentation is crucial for disease mechanism assessment and drug discovery.
• Existing methods often rely on image binarization, which can be sensitive to smoothing parameters.
• Inappropriate Gaussian smoothing in histogram thresholding can lead to inaccurate cell segmentation.

Purpose of the Study:

• To develop an improved cell segmentation technique for fluorescent microscopy images.
• To address the limitations of traditional histogram thresholding methods.
• To enhance the accuracy of cell segmentation, particularly for overlapping cells.

Main Methods:

• A novel multi-scale histogram thresholding (MHT) technique is proposed.
• The MHT method involves smoothing image histograms at multiple scales (Gaussian standard deviations).
• Smoothed histograms are fused, and thresholding is applied for binarization, integrated into a framework with region-based ellipse fitting for overlapping cell identification.

Main Results:

• The proposed MHT technique demonstrates superior performance in cell segmentation compared to existing methods.
• Experimental results on benchmark datasets validate the effectiveness of the MHT approach.
• The integrated framework successfully improves segmentation accuracy and handles overlapping cells.

Conclusions:

• The multi-scale histogram thresholding technique offers a robust solution for accurate cell segmentation.
• This method enhances the reliability of image analysis in biological research.
• The approach provides a significant advancement for applications in disease mechanism assessment and drug discovery.