Recognition of parasitic helminth eggs via a deep learning-based platform

Area of Science:

• Medical Diagnostics
• Parasitology
• Artificial Intelligence

Background:

• Traditional methods for diagnosing parasitosis, like microscopy, are time-consuming and prone to errors.
• Accurate and rapid diagnosis is essential for effective treatment of parasitic infections.

Purpose of the Study:

• To evaluate the use of the YOLOv4 deep learning algorithm for detecting and classifying human parasite eggs.
• To develop an AI-assisted platform for improved diagnosis of parasitosis.

Main Methods:

• Collected eggs from various human parasite species, including Ascaris lumbricoides, Trichuris trichiura, and Schistosoma japonicum.
• Prepared single and mixed egg smears, photographed them under a light microscope, and analyzed them using the YOLOv4 object detection model.

Main Results:

• The YOLOv4 model achieved 100% accuracy for Clonorchis sinensis and Schistosoma japonicum.
• Recognition accuracies for other species ranged from 84.85% to 89.31%, with high accuracies for mixed egg samples (up to 98.10%).

Conclusions:

• The AI-assisted platform demonstrates high accuracy and efficiency in detecting and classifying parasite eggs.
• This technology can reduce dependence on expert knowledge and enhance parasitosis diagnosis and treatment efforts.