Entomopathogenic nematode detection and counting model developed based on A-star algorithm

Area of Science:

• Agricultural Science
• Nematology
• Computer Vision

Background:

• Entomopathogenic nematodes are crucial for biological pest control, offering an alternative to chemical pesticides.
• Manual counting of these nematodes in laboratory settings is laborious, time-consuming, and prone to approximation.
• Accurate quantification is essential for effective application and research in biological control.

Purpose of the Study:

• To develop and evaluate a novel computer vision-based method for the detection and quantification of Steinernema feltiae.
• To improve the efficiency and accuracy of nematode counting in laboratory studies.
• To provide a viable automated alternative to traditional manual counting methods.

Main Methods:

• A computer vision approach utilizing an A-star-based network for nematode detection and isolation from microscope images.
• The algorithm focuses on framing and isolation steps, optimizing for speed and space complexity.
• Performance was compared against established YOLO models (YOLO-V5m, YOLO-V7m, YOLO-V8m).

Main Results:

• The developed A-star-based network significantly outperformed YOLO-V5m, YOLO-V7m, and YOLO-V8m in detection accuracy.
• The method demonstrated high efficacy in detecting overlapping entomopathogenic nematodes.
• The algorithm's design minimized computational overhead, excluding complex processes like weight documents and model integration.

Conclusions:

• The proposed computer vision method is feasible and effective for detecting and counting entomopathogenic nematodes.
• This automated approach offers a more efficient and accurate solution for nematode quantification in research and application.
• The study highlights the potential of AI-driven tools in advancing biological control strategies.