Trends and advances in image-based mosquito identification and classification using machine learning models: A...

Area of Science:

• Entomology
• Computer Science
• Public Health

Background:

• Mosquito-borne diseases (e.g., Yellow fever, Dengue, Zika) are a major global health concern.
• Traditional mosquito identification is slow, labor-intensive, and requires expert knowledge.
• Machine learning (ML) offers automated, rapid, and accurate species identification.

Purpose of the Study:

• To systematically review ML-based mosquito identification research.
• To evaluate the strengths, limitations, and geographic contributions of ML in this field.
• To identify strategies for improving ML model generalizability and applicability in vector control.

Main Methods:

• Systematic literature review of 52 studies (2000-2024) from major academic databases.
• Analysis focused on ML techniques, feature extraction, classification accuracy, and identified limitations.
• Inclusion criteria selected from 1,050 initial papers.

Main Results:

• ML models achieve high accuracy using morphological features (wings, body structures) for species, sex, and age classification.
• Feature extraction is crucial for capturing fine-grained traits.
• Key limitations include dataset diversity, inconsistent preprocessing, high computational needs, and inter-species morphological similarities.

Conclusions:

• ML shows significant potential to enhance mosquito surveillance and vector control strategies.
• Addressing dataset limitations, computational requirements, and deployment strategies is crucial for real-world impact.
• Promoting diverse datasets and African-led research can foster inclusive and context-relevant systems.