Assessment of Novel Acaricides to Counter European Honeybee Memory Loss and Varroa Destructor Resistance using Image Recognition and Artificial Intelligence Approach

An incredibly resilient and destructive pest, Varroa destructor has proved to be the greatest biological threat to the Western honeybee and the bee-based economy as a whole. Current methods of controlling varroa mite populations are becoming increasingly insufficient, leading to the necessity for more effective acaricides. Aiding the effort for efficient testing of acaricides, we developed the first AI-based semantic segmentation model to track and analyze a bee's full flight path. Named BeeTrack, this model can monitor bee decision-making processes and bee behavior quickly and cost-effectively. Using BeeTrack and our training setup and methods, we evaluated commercial and experimental acaricides and their effects on bee memory. The memory training utilized three quinine-coated yellow and three sucrose-coated blue landing platforms. A blanket covered the training box and LEDs were stored in the platforms to reduce phototaxis. After 2 training trials, 45 bees in total were exposed to each chemical and continued to the testing phase, where platforms contained water, and the resulting footage of bee landing patterns was analyzed. Lactic acid (LA), 2 essential oil mixtures (EO1 and EO2), and capric acid (CA) were our experimental chemicals, with untreated bees and commercial acaricides Apiguard and FormicPro being our controls. BeeTrack analyzed each bee's direction, speed, and path. Untreated bees performed with 81.82% landing accuracy, measured by their landings on blue platforms over their total landings. Of our chemicals, EO2-exposed bees performed the best (75% accuracy), followed by EO1 (71.4%), and CA (70%), outperforming FormicPro and Apiguard (45%, 55.5%). Thus, we have designed a free-flying training environment and procedure to train bees with a minimal number of trials. Using our BeeTrack system, we were able to analyze high-speed bee flight paths while bypassing much of the manual analysis and find chemicals with high potential as acaricides.