Image processing and classification algorithm for yeast cell morphology in a microfluidic chip

Area of Science:

• Computational Biology
• Microscopy Image Analysis
• Yeast Cell Biology

Background:

• Accurate yeast cell cycle phase identification is crucial for morphological studies.
• Manual classification of yeast cell morphology is time-consuming and subjective.
• Microfluidic environments present unique challenges for image analysis.

Purpose of the Study:

• To develop and evaluate a computer-based algorithm for automatic classification of yeast cell cycle phases.
• To extract and utilize morphological features like bud size for classification.
• To assess the performance of machine learning classifiers under varying image conditions.

Main Methods:

• Image enhancement techniques to reduce background noise.
• Development of a robust segmentation algorithm to extract geometrical features (compactness, axis ratio, bud size).
• Comparison of linear support vector machine, distance-based classification, and k-nearest-neighbor algorithms for classification.

Main Results:

• The algorithm successfully extracts key morphological features from yeast cell images.
• Machine learning classifiers demonstrated effectiveness in classifying yeast cell cycle phases.
• The system showed robust performance despite variations in illumination and focus.

Conclusions:

• Automated classification of yeast cells based on morphology is feasible.
• The developed algorithm can accurately classify yeast cells even with noisy and low-contrast images.
• This approach offers a consistent and efficient method for yeast cell cycle analysis.