Counting nematodes made easy: leveraging AI-powered automation for enhanced efficiency and precision

Area of Science:

• Agricultural Science
• Nematology
• Machine Learning

Background:

• Nematode counting is crucial for quantitative studies but is labor-intensive and time-consuming.
• Existing automated methods focus on single-class object identification, limiting practical application.
• There is a need for algorithms capable of concurrent detection and classification of multiple nematode types.

Purpose of the Study:

• To develop a user-friendly Graphical User Interface (GUI) for automated nematode counting.
• To enable simultaneous recognition and categorization of multiple nematode classes, specifically Meloidogyne spp. eggs and juveniles.
• To leverage deep learning models for enhanced accuracy and efficiency in nematode population density assessments.

Main Methods:

• Generated 650 images of nematode eggs and 1339 images of Meloidogyne juveniles using two imaging systems.
• Annotated 8655 eggs and 4742 juveniles using bounding box and segmentation techniques.
• Developed deep learning models utilizing the YOLOv8x Convolutional Neural Networks (CNNs) architecture.

Main Results:

• The models achieved 94% accuracy in identifying eggs and 93% accuracy in identifying Meloidogyne juveniles.
• Demonstrated a coefficient correlation greater than 0.70 between model predictions and observations on unseen images.
• The developed GUI, incorporating these models, is publicly available on GitHub.

Conclusions:

• The study successfully demonstrated the potential utility of deep learning models for practical, automated nematode counting.
• The developed GUI offers a user-friendly solution for simultaneous recognition and categorization of multiple nematode types.
• This work serves as a foundation for a universal automated nematode counting system, encouraging global collaboration.