Nuclear boundary detection algorithm based on a minimax derivative statistic for atypical bronchial squamous...

Area of Science:

• Digital image analysis
• Cell biology
• Biomedical imaging

Background:

• Traditional thresholding methods are unreliable for segmenting squamous epithelial cells due to nuclear chromatin condensation.
• Nuclear boundaries in these cells are visually distinct despite challenges in automated detection.

Purpose of the Study:

• To develop and present an algorithm for accurate nuclear boundary detection in digitized cell images.
• To address the unreliability of thresholding in specific cell types.

Main Methods:

• Developed a novel algorithm utilizing a minimax derivative statistic to identify nuclear boundaries.
• The statistic exhibits maximum values at boundaries and low values elsewhere.
• Outlined statistical properties of the confusion matrix for cellular scene segmentation.

Main Results:

• The minimax derivative statistic effectively detects nuclear boundaries in digitized cell images.
• Presented a measurement and test statistic for quantifying scene segmentation errors.
• The new algorithm offers improved reliability over thresholding for specific cell types.

Conclusions:

• The developed algorithm provides a more reliable method for nuclear boundary detection in squamous epithelial cells.
• This advancement can enhance the accuracy of cell image segmentation and analysis.
• Further statistical analysis of segmentation errors is presented.