Edge Artificial Intelligence (AI) for real-time automatic quantification of filariasis in mobile microscopy

Area of Science:

• Medical Parasitology
• Artificial Intelligence in Healthcare
• Public Health Technology

Background:

• Filariasis, a neglected tropical disease, poses a significant public health challenge in tropical regions.
• Current diagnostic methods like microscopy are time-consuming and require specialized expertise, limiting accessibility.
• Artificial intelligence (AI) offers a potential solution for automated detection and differentiation of microfilariae.

Purpose of the Study:

• To develop and validate an edge AI system for real-time filariasis diagnosis on a smartphone.
• To meet the World Health Organization's target product profile for lymphatic filariasis diagnosis.
• To enable accurate detection and species differentiation of microfilariae without internet connectivity.

Main Methods:

• An object detection algorithm (SSD MobileNet V2) was trained and validated using microscopic blood sample data.
• An edge AI system was developed, integrating the algorithm into a smartphone camera attached to an optical microscope.
• The system was clinically validated in a real-world setting using augmented microscopy.

Main Results:

• The AI system successfully detected microfilariae at 10x magnification and differentiated four species (Loa loa, Mansonella perstans, Wuchereria bancrofti, Brugia malayi) at 40x magnification.
• The screening algorithm achieved 94.14% precision, 91.90% recall, and 93.01% F1 score.
• The species differentiation algorithm achieved 95.46% precision, 97.81% recall, and 96.62% F1 score.

Conclusions:

• The developed smartphone-based edge AI system provides an innovative solution for filariasis diagnosis.
• This technology can significantly support filariasis diagnosis and monitoring, especially in resource-limited settings.
• The system offers real-time, accurate, and accessible diagnostic capabilities for neglected tropical diseases.