Superior Auto-Identification of Trypanosome Parasites by Using a Hybrid Deep-Learning Model

Area of Science:

• Parasitology
• Medical Entomology
• Artificial Intelligence in Healthcare

Background:

• Trypanosomiasis poses a significant global public health challenge, particularly in South and Southeast Asia.
• Effective disease control relies on identifying transmission hotspots through active surveillance.
• Current diagnostic methods, like microscopy, require skilled personnel, limiting widespread application.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI) program for automated identification and classification of trypanosome species.
• To assess the AI program's performance using comprehensive statistical metrics and diagnostic curves.
• To provide a rapid, accurate screening tool to aid in trypanosomiasis surveillance and control.

Main Methods:

• A hybrid deep learning approach combining object identification and classification neural networks was implemented on the CiRA CORE AI platform.
• The AI program analyzes oil-immersion microscopic images to detect and differentiate Trypanosoma cruzi, T. brucei, and T. evansi.
• Attention maps highlight key parasite features (nucleus, kinetoplast) for AI analysis.

Main Results:

• The AI program demonstrated effectiveness in identifying and categorizing trypanosome parasites from microscopic images.
• Performance was rigorously assessed using modules generating metrics like accuracy, precision, recall, F1 score, ROC, and PR curves.
• The AI algorithm showed high accuracy in parasite detection and classification.

Conclusions:

• The developed AI program offers a promising solution for automated and accurate trypanosome species identification.
• This technology can significantly enhance disease surveillance, enabling faster and more informed decision-making for control strategies.
• The AI tool has the potential to revolutionize public health efforts against trypanosomiasis.