Diatom identification including life cycle stages through morphological and texture descriptors

Area of Science:

• Microbiology
• Algology
• Computational Biology

Background:

• Diatoms are vital unicellular algae found globally in aquatic environments.
• Diatom identification is crucial for ecology, forensic science, and water quality assessment.
• Modeling diatom life cycles is complex due to non-linear, high-dimensional variations.

Purpose of the Study:

• To develop a method for characterizing diatom life cycles.
• To identify key features that change throughout the algae life cycle, focusing on shape and texture.

Main Methods:

• Utilized Elliptical Fourier Descriptors (EFD) for diatom contour analysis.
• Employed phase congruency and Gabor filters to analyze inner diatom ornamentation.
• Applied supervised and non-supervised classification techniques to an algae dataset.

Main Results:

• Achieved high accuracy in diatom classification: up to 99% with supervised learning and 98% with non-supervised learning.
• Successfully characterized diatom life cycles based on shape and texture features.
• Demonstrated the efficacy of EFD, phase congruency, and Gabor filters in diatom analysis.

Conclusions:

• The proposed method effectively characterizes diatom life cycles using morphological and textural features.
• This approach offers a robust tool for diatom identification and analysis in various scientific fields.
• High classification accuracies indicate the potential of this method for ecological and forensic applications.