The Automatic Classification of Pyriproxyfen-Affected Mosquito Ovaries

Area of Science:

• Entomology
• Bioinformatics
• Insecticide Resistance

Background:

• Pyrethroid-resistant vectors pose a significant challenge to malaria control.
• Pyriproxyfen (PPF) is a potential alternative insecticide for controlling resistant mosquito populations.
• Assessing PPF efficacy traditionally involves expert-driven dissection and analysis of vector ovaries, which is time-consuming and requires specialized skills.

Purpose of the Study:

• To develop and validate a convolutional neural network (CNN) for automated classification of mosquito egg development and fertility status.
• To overcome the limitations of manual ovary dissection and assessment for evaluating PPF efficacy.
• To provide a rapid, robust, and accessible method for assessing insecticide effectiveness.

Main Methods:

• A ResNet-50 CNN was pretrained on the ImageNet dataset and retrained using a novel dataset of 524 dissected ovary images from An. gambiae s.l., An. gambiae Akron, and An. funestus s.l.
• The dataset included images with known fertility status and PPF exposure.
• Data augmentation techniques were employed to expand the training set to 6973 images, and a test set of 157 images was used for validation.

Main Results:

• The developed CNN model achieved a high accuracy score of 94% in classifying fertility status.
• The automated analysis using the CNN took a mean time of 38.5 seconds per assessment.
• The model demonstrated a practical, accessible, and free-to-distribute format for widespread use.

Conclusions:

• The CNN approach offers a precise, quick, and robust method for assessing PPF efficacy, overcoming the limitations of manual methods.
• This automated technique is valuable for evaluating the effectiveness and durability of PPF-treated tools, including bednets.
• The methodology is adaptable for assessing other insecticides with similar modes of action and various vector control applications.