Validation of supervised machine learning Vectrack algorithms for automatic counting and identification of mosquito adults and exploitation for studying their circadian rhythms

Background: Mosquito-borne diseases are an increasing public health concern, also in Italy, where West Nile virus is endemic, and risk of exotic arbovirus transmission is rising, as highlighted by the two Dengue outbreaks occurred in 2023 and an even larger one in 2024. Effective monitoring of disease-vector mosquitoes is crucial to optimize control interventions and make evidence-based risk assessments. However, traditional entomological monitoring methods are labor-intensive, time-consuming and lack high temporal and spatial resolution. As a result, there is a growing interest in developing innovative approaches, requiring lower human efforts, such as automated mosquito counting and identification tools. In 2022, a novel system combining an optical sensor with machine learning technology (VECTRACK) proved effective in counting and identifying Aedes albopictus and Culex pipiens adult females and males. Here, we conducted the first large-scale field evaluation of the VECTRACK system to assess: (i) whether the catching capacity of a commercial BG-Mosquitaire trap (BGM) for adult mosquito equipped with VECTRACK (BGM+VECT) was affected by the sensor; (ii) the accuracy of the VECTRACK algorithm in classifying the target mosquito species genus and sex. Methods: The same experimental design was implemented in four areas across Italy in July-September 2023: northern (Bergamo and Padua districts), central (Rome), and southern (Procida Island, Naples). In each area, three trap types — one BGM, one BGM+VECT and the combination of four sticky traps (STs) — were rotated every 48 hours across three different sites. Each sampling scheme was replicated three times/area. Collected mosquitoes were counted and identified both by the VECTRACK algorithm and operatormediated morphological examination. The VECTRACK system’s performance was evaluated using generalized linear mixed and linear regression models. Results: A total of 3,829 mosquitoes (90.2% Ae. albopictus) were captured in 18 collection-days/trap/site. BGM and BGM+VECT showed comparable performance in collecting target mosquitoes. Results showed a higher correlation between visual and automated identification methods for Ae. albopictus (Spearman: females = 0.97; males = 0.89; P < 0.0001), than for Cx. pipiens (Spearman: females = 0.67, P < 0.0001; males = 0.37; P = 0.025), with minimal counting errors. Results of the linear model estimate that for every 100 individuals of each species identified by the Sensor, the traps had on average actually captured 105 Ae. albopictus and 77 Cx. pipiens, respectively (Chisq NS). Conclusions: Results strongly support the VECTRACK system as a powerful tool for Ae. albopictus and Cx. pipiens research and its high potential for continuous monitoring with minimal human effort. In addition, the recording of the exact time of each mosquito capture opens the unprecedented possibility to study mosquito circadian rhythms under natural conditions. To this goal, nine institutions within Research Node 2 of the INF-ACT project are exploiting the VECTRACK system to carry out a largescale study of the circadian rhythms of Ae. albopictus and Cx. pipiens in urban/semi-urban areas across Italy in the reproductive season 2025.