Species identification of phlebotomine sandflies using deep learning and wing interferential pattern (WIP)

Area of Science:

• Medical Entomology
• Bioinformatics
• Taxonomy

Background:

• Sandflies are critical vectors of various pathogens, posing significant public health and veterinary risks.
• Accurate sandfly identification is crucial for disease surveillance and control but is often hindered by complex morphological examinations.

Purpose of the Study:

• To evaluate the efficacy of Wing Interferential Patterns (WIPs) and deep learning (DL) for accurate sandfly identification.
• To establish a non-invasive method for sandfly classification at multiple taxonomic levels.

Main Methods:

• Generating Wing Interferential Patterns (WIPs) on sandfly wings.
• Applying deep learning (DL) algorithms to analyze WIPs for taxonomic assignment.
• Validating the method across family, genus, subgenus, and species levels.

Main Results:

• The WIPs-DL method achieved identification accuracy exceeding 77.0% across all tested taxonomic levels.
• The approach successfully distinguished sandflies from other dipteran insects.
• Identification did not require examination of internal structures or traditional keys.

Conclusions:

• Wing Interferential Patterns (WIPs) coupled with deep learning (DL) offer a reliable and efficient method for sandfly identification.
• This technique is suitable for field conditions, supporting proactive and passive entomological surveys.
• The approach addresses the challenge of identifying sandflies, especially amid a shortage of medical entomologists.