Skip to main content

Automated classification of mixed populations of Aedes aegypti and Culex quinquefasciatus mosquitoes under field conditions

Abstract

Background

The recent rise in the transmission of mosquito-borne diseases such as dengue virus (DENV), Zika (ZIKV), chikungunya (CHIKV), Oropouche (OROV), and West Nile (WNV) is a major concern for public health managers worldwide. Emerging technologies for automated remote mosquito classification can be supplemented to improve surveillance systems and provide valuable information regarding mosquito vector catches in real time.

Methods

We coupled an optical sensor to the entrance of a standard mosquito suction trap (BG-Mosquitaire) to record 9151 insect flights in two Brazilian cities: Rio de Janeiro and Brasilia. The traps and sensors remained in the field for approximately 1 year. A total of 1383 mosquito flights were recorded from the target species: Aedes aegypti and Culex quinquefasciatus. Mosquito classification was based on previous models developed and trained using European populations of Aedes albopictus and Culex pipiens.

Results

The VECTRACK sensor was able to discriminate the target mosquitoes (Aedes and Culex genera) from non-target insects with an accuracy of 99.8%. Considering only mosquito vectors, the classification between Aedes and Culex achieved an accuracy of 93.7%. The sex classification worked better for Cx. quinquefasciatus (accuracy: 95%; specificity: 95.3%) than for Ae. aegypti (accuracy: 92.1%; specificity: 88.4%).

Conclusions

The data reported herein show high accuracy, sensitivity, specificity and precision of an automated optical sensor in classifying target mosquito species, genus and sex. Similar results were obtained in two different Brazilian cities, suggesting high reliability of our findings. Surprisingly, the model developed for European populations of Ae. albopictus worked well for Brazilian Ae. aegypti populations, and the model developed and trained for Cx. pipiens was able to classify Brazilian Cx. quinquefasciatus populations. Our findings suggest this optical sensor can be integrated into mosquito surveillance methods and generate accurate automatic real-time monitoring of medically relevant mosquito species.

Graphical Abstract

Background

Emerging and re-emerging vector-borne diseases (VBD) represent significant global public health concerns, showing a rising incidence rate in endemic areas in the last decades [1, 2]. Equally important, VBD have expanded into new regions following an increase in the geographic distribution of their primary vectors [3]. Most of the VBD impacting public health systems are viruses and are transmitted by Aedes and Culex mosquitoes. Special interest is given to Aedes aegypti, Ae. albopictus, Culex quinquefasciatus and Cx. pipiens, species that are well adapted to living in association with humans. Aedes aegypti and Cx. quinquefasciatus are more abundant in urbanized areas and collected more frequently in the intradomestic environment of tropical regions. Aedes albopictus is more eclectic in biting behavior and habitat preferences, being found from suburban to peridomestic areas with high vegetation coverage [4,5,6,7,8,9,10,11].

Several biotic and abiotic factors can be linked to the recent expansion of VBD. Temperature seems to be the most limiting factor regarding vector distribution and thus disease transmission to temperate regions. Considering climate change and the expected increase in temperature in the next years, a subject for research is to foresee the impact of temperature rise on the transmission of arboviruses such as dengue (DENV), Zika (ZIKV), chikungunya (CHIKV), mayaro (MAYV), West Nile (WNV) and St. Louis encephalitis (SLE), to mention a few diseases notably transmitted by the three aforementioned mosquito species [12,13,14,15,16,17]. Climate change's impact is particularly significant in the twenty-first century, with some studies projecting a global temperature increase of 1.0–3.5 °C by 2100, intensifying arbovirus transmission but also expanding their occurrence to areas normally free from the viruses or with mild incidence [3]. Among the biotic factors influencing arbovirus transmission, human activities such as unplanned urbanization, construction of dams and irrigation schemes, increased global travel and trade, deforestation and the spread of insecticide resistance are factors playing a role in shaping mosquito and disease distribution [18,19,20,21,22].

We need to build and strengthen early warning systems and increase response capacities and preparedness for current and future threats. Regarding mosquito vectors, one of the greatest research gaps involves the implementation of a surveillance agenda that can efficiently determine seasonal mosquito population fluctuation and identify the high-risk areas in endemic metropolitan cities [23,24,25]. Undoubtedly, considering arbovirus epidemiology and disease dynamics, surveillance needs to determine the spatiotemporal risk in a timely manner, allowing further intervention (e.g. vector control activities such as ULV fogging, removal of potential breeding sites, community engagement, urban cleaning, etc.) to mitigate or halt transmission [26,27,28]. To estimate the sub-areas of a city with higher risk of disease transmission, vector sampling must be conducted citywide [26]. Regarding the biology of Aedes and Culex mosquitoes, performing mosquito sampling in the larval stage is unfeasible because of the many health agents required to sample every premise, the tedious nature of the work, which may impact the search effort, and a lack of long-term paramilitary organization to conduct larval surveys [23, 29]. Therefore, use of mosquito traps is recommended to make citywide sampling feasible. Additionally, it is well known that although immature stage monitoring might be simpler to set up, it lacks utility in estimating adult abundance and transmission risk due to poor correlation with egg, larval and pupal density indices [23, 25, 30]. A growing body of evidence shows that monitoring mosquito populations with adult traps better represents seasonal variations in mosquito abundance compared to indices derived from immature stage-based sampling like the House or Breteau Index [23, 30,31,32,33]. Myriad mosquito traps are available using different types of attractants and mechanisms, with their pros and cons [34]. Conventional traps with catch bags require periodic field collection and time-consuming entomological identification by an expert holding an appropriate taxonomic key, leading to delays in analyzing mosquito population dynamics [35]. Therefore, new approaches such as optical sensors combined with machine learning could provide almost real-time mosquito classification to support surveillance programs with timely determination of mosquito composition as soon as target species are trapped [36]. While historically wingbeat frequency has been the primary predictor variable, recent efforts have focused on improving classification methods to distinguish mosquito species, sex and even parity status [36]. Locally acquired field data on mosquito captures could be modeled with climatic records, traps, spatiotemporal localization and ecological features not only to improve remote mosquito classification in the field but also to develop real-time wireless entomological surveillance.

The development of automatic devices to further classify mosquito species is a growing field. It is important to test those devices under field conditions to provide a realistic challenge and reliable estimates. We report here an independent evaluation of a prototype optical sensor coupled to the BG-Sentinel [37, 38], a commercial mosquito trap commonly used in the field for trapping urban mosquito vectors. This evaluation included field sampling of Aedes and Culex mosquitoes in two sites in Brazil with subsequent adoption of previously established models to determine mosquito counts [36]. Herein, we report the effectiveness of the VECTRACK sensor in identifying and counting Ae. aegypti and Cx. quinquefasciatus in Rio de Janeiro and Brasilia using a classification model developed for Ae. aegypti and Cx. pipiens in Europe [36], highlighting the relevance of this model for genus classification.

Methods

Sensor and trap description

The VECTRACK sensor, developed and manufactured by Irideon SL in Barcelona, Spain, was integrated with the entrance of a commercial BG-Mosquitaire suction trap from Biogents AG in Regensburg, Germany, as depicted in Fig. 1. Under a field setting, the BG-Mosquitaire trap captures mostly host-seeking mosquito females that are attracted to both visual and sensory cues, like its predecessor, BG-Sentinel [37, 38]. This trap is equally efficient in trapping both target species in field scenarios with high and low mosquito population density [37, 38]. The trap was equipped with a Biogents AG BGSweetscent chemical attractant sachet, and the fan-driven airflow within the trap was maintained at approximately 3 m/s in the downward direction. Mosquitoes flying in proximity to the sensor’s entrance funnel could be drawn in by the fan, detected by the sensor, and subsequently trapped in the catch bag within the trap body.

Fig. 1
figure 1

Time series plots representing the number of target mosquitoes (sensor count and manual count) per field assay. The x-axis indicates the end date of each field assay. A Sensor 1, B Sensor 2, C data gathered in Rio de Janeiro, D data gathered in Brasilia

The sensor itself comprises an optical emitter panel and an optical receiver panel, positioned facing each other through a transparent flight tube with 105 mm diameter. The optical emitter consists of a two-dimensional (2D) array of 940-nm wavelength infrared light-emitting diodes (LEDs), while the optical receiver incorporates a 2D array of 940-nm photodiodes. The active length of the optical sensor in the downward direction is 70 mm. The output from the optical receiver undergoes amplification and is acquired by an analog-to-digital converter (ADC) with a sampling frequency of 9603 samples per second. Upon a mosquito entering the sensing volume, it triggers an automatic recording of up to 1024 samples, with a duration of up to 107 ms. Considering the typical duration of a mosquito’s flight is approximately 50 ms, this setup enables effective monitoring. Additionally, the sensor automatically timestamps each recording and captures the measured ambient temperature.

In this study, we targeted Ae. aegypti and Cx. quinquefasciatus from two field sites from Brazil, as described below. The classification of field-gathered specimens was based on previously developed and trained models. For Ae. aegypti, we used a model that was developed using both Spanish and Portuguese populations of Ae. albopictus, originally collected from Rubi, Barcelona (41°29′49″N; 02°02′05″E), and Algarve, Portugal (37°10′48″N; 08°22′22″E) [39]. The model used to classify Brazilian populations of Cx. quinquefasciatus was developed and trained for Spanish populations of Cx. pipiens obtained from Cerdanyola del Vallés, Barcelona (41°29′10″N; 02°04′29″E) [39].

Data acquisition process

The two available VECTRACK sensors were exposed to a total of 28 non-simultaneous assays between July 2022 and May 2023. From those 28 assays, 21 were done in Rio de Janeiro and 7 in Brasilia. The two sensors were rotated in both cities to avoid any bias possibly related to specific geographic condition. Each assay consisted of exposing the VECTRACK sensor and BG-Mosquitaire to the field conditions reported above for a period ranging from 12 to 15 consecutive days.

Study area and field trial conditions

The sensors were installed in two Brazilian cities: Rio de Janeiro and Brasilia. In Rio de Janeiro, the sensors were installed in a commercial company in Jacarepaguá (22°58′35″S; 43°24′59″W), a neighborhood characterized by vegetation coverage, mild mean temperature and presence of Ae. albopictus [7]. In Brasilia, the sensors were installed at the National Health Surveillance Department (15°48′53″S; 47°54′59″W), a highly urbanized area with almost no vegetation coverage. In each field site, the two sensors were installed indoors and outdoors to capture the diversity of mosquito vectors in the surroundings. Traps remained in the field for periods between 12 and 15 days, and all assays were conducted in an approximately 1 year period. The catch bags of the BG-Mosquitaire were inspected twice per week, i.e. around four times per assay. Traps remained plugged into power during the assays. The insects were identified using a Zeiss stereomicroscope and recorded considering the sensor and its location. The taxonomic identification of mosquito vectors was conducted using the appropriate local entomological keys [40]. A taxonomic identification of non-target species up to species level was out of the scope of this article.

Data analysis of sensor classification in the field

Two main functions of the automated sensor were assessed: (i) the ability of the system to discriminate Ae. aegypti and Cx. quinquefasciatus target mosquitoes from non-target insects that were accidentally trapped and (ii) the ability of the system to correctly discriminate target mosquito species' genus and sex. The potential relationship between sensor count (mosquitoes counted by the sensor) and manual count (mosquitoes counted by manual inspection) was evaluated through correlation and linear regression analyses, with the results depicted using both time series and scatter plots for each collection cycle.

Pearson correlation coefficient (r) and associated P-values were calculated to assess the strength and significance of the relationship between the two variables. Additionally, regression coefficients, including the coefficient of determination (R2), linear slope and intercept, were computed to evaluate how well sensor count predictions aligned with manual counts. A regression slope greater or less than one indicated whether sensor counts tended to be higher or lower, respectively, than manual counts on average.

We used confusion matrix to evaluate the accuracy, sensitivity, specificity and precision of the classification tasks proposed. One confusion matrix was developed for each classification task, which involved classifying (i) target or non-target species, (ii) Aedes or Culex, (iii) Ae. aegypti sex and (iv) Cx. quinquefasciatus sex. For each confusion matrix, we determined the number of true positives (TP), true negatives (TN), false positives (FP) and false negatives (FN) using the manual classification as the reference gold standard. TP and TN are the numbers of positive and negative cases respectively that the system classified correctly. FP is the number of negative cases that the system incorrectly classified as positive, and FN is the number of positive cases that the system incorrectly classified as negative. To calculate TP, TN, FP and FN for a particular class, this class was defined as the positive class and the other class(es) were defined as the negatives. For the positive class, TP equals the minimum common value of the sensor and manual count. If the sensor count was greater than the manual count, then the difference was taken as FP; otherwise, FP equaled zero. If the sensor count was less than the manual count, then the difference was taken as FN. TN is calculated by subtracting FP from the manual counts for the negatives. Accuracy means how often the classifier is correct and was measured as (TP + TN)/total samples. Sensitivity, also known as true-positive rate or recall, indicates the proportion of positives that are correctly classified by the system and is represented as TP/(TP + FN). Specificity, also known as true-negative rate, indicates the proportion of negatives that are correctly classified by the system and is estimated TN/(TN + FP). Finally, the precision evaluates when the sensor predicts the sample as positive, how often it is correct. Precision is determined by TP/(TP + FP).

Results

Manual target species identification

A total of 1300 mosquitoes were sampled in all 28 assays, resulting in an average of 46.4 mosquitoes trapped per assay. From the total mosquitoes sampled, 627 (48.2%) were Cx. quinquefasciatus (545 females and 82 males) and 662 (50.9%) Ae. aegypti (436 females and 226 males). We still sampled 11 Ae. albopictus (10 females and 1 male), corresponding to 0.8% of the sampled specimens. Regarding the two field sites, 1215 (93.5%) mosquitoes were captured in Rio de Janeiro and 159 in Brasilia (12.2%). Focusing on the target species, in Rio de Janeiro, 43.7 and 55.4% were Ae. aegypti and Cx. quinquefasciatus, whereas in Brasília these species represented 82.4 and 17.6%, respectively.

Manual non-target insect species identification

During all the 28 field assays, a total of 8012 non-target insects were sampled. From this sum, 6826 (85.2%) were sampled in Rio de Janeiro, whereas 1186 (14.8%) were collected in Brasilia. Among the non-target sampled insects, most insects were flies and other small dipterans, although we did not conduct the appropriate taxonomic identification of non-targeted specimens up to species level.

Automated detection of target mosquito species in the field

A remarkable overlapping between manual and sensor counts was observed in the two sensors used in the 28 field assays in both cities (Fig. 1). Linear regression analysis indicated a good fit of the linear regression line to manual count versus sensor count (Fig. 2). However, we should highlight the fit varied according to the species*genus combination. The coefficient of determination was > 0.99 in all cases, with the exception of males of Cx. quinquefasciatus, with a R2 equaling 0.889. Although we still consider it a satisfactory fit, further investigations should understand the hurdles in classifying Cx. quinquefasciatus males. Taking all the data together, i.e. using data from both sensors, both target species (Ae. aegypti and Cx. quinquefasciatus), plus males and females, a final coefficient of determination of 0.991 was observed, suggesting the VECTRACK sensor performed well in identifying mosquitoes.

Fig. 2
figure 2

Scatter plot and linear regression of sensor count versus manual count for A Aedes aegypti and B Culex quinquefasciatus per mosquito gender showing the regression line equation (slope and y-intercept) and coefficient of determination. In A, the regression line for female and male is y = 1.0706x–0.1224 and y = 1.1085x + 0.1879, whereas the coefficient of determination is R2 = 0.992 and R2 = 0.9819 for female and male, respectively. In B, the regression line for female and male is y = 1.066x–0.248 and y = 0.9023x + 0.1432, whereas the coefficient of determination is R2 = 0.995 and R2 = 0.8897 for female and male, respectively

Confusion matrix and classification parameters

We developed one confusion matrix for each classification task. The most impressive results were obtained for classifying samples between target or non-target species, with accuracy, sensitivity, specificity and precision > 0.99 (Table 1). Those identified as target species were later classified under the genus Aedes or Culex. In this case, the VECTRACK sensor was able to accurately classify samples in 93.7% of the cases. In the genus classification, we observed the lowest sensitivity of the classification tasks we proposed: 92.1%. It indicates that around 8% of the samples were FN, i.e. the sensor classified them as Culex but they were Aedes after manual check. Regarding sex classification, better results were obtained for Culex, although similar results in sensitivity were observed for male and female sorting for both genera (Table 1).

Table 1 Summary of classification parameters based on confusion matrixes developed accordingly to the classification task

Discussion

The development of automated identification sensors and systems designed for accurately identifying medically relevant mosquito species has seen significant advancement in recent years [41,42,43]. These ‘smart traps’ aim to provide reliable information regarding the identification of target species, ideally sorting them by sex. This is particularly important as it allows for the timely detection of female mosquitoes, a crucial parameter for estimating local disease transmission risk. In our study, we recorded the flight patterns of 10,534 insects in two distinct field sites in Brazil, characterized by different landscapes, abiotic conditions and arbovirus epidemiology. Among these, 1383 flight records belonged to medically important mosquitoes. To our knowledge, this report represents the first instance of utilizing sensors to classify target mosquito species and their sex within a dengue-endemic country.

Brazil has been experiencing over 1 million dengue cases annually since the late 2010s. By the end of April 2024, > 3 million dengue cases had been reported in the country, with concurrent circulation of CHIKV and ZIKV, albeit in smaller proportions [44, 45]. In this context, reliable tools for rapidly classifying field-caught mosquitoes are essential for public health management in endemic areas. Immediate identification of disease vectors can promote timely interventions. For instance, Brazil has been investing in strengthening its surveillance capacity for arboviruses. The city of Foz do Iguaçu has developed an innovative surveillance network based on One Health principles, focusing on five main areas: (i) integrating previously sectorized field teams into a unified One Health team capable of performing multiple tasks after receiving household allowances for indoor inspections; (ii) adopting digital solutions to replace archaic practices such as recording field information on printed spreadsheets; (iii) empowering health agents and providing them with continuous training alongside local scientists; (iv) mobilizing communities through meetings at schools, churches and community centers; (v) conducting active surveys to gain a better understanding of dengue epidemiology [27, 46]. A cornerstone of this surveillance system involves implementing adult mosquito sampling through extensive citywide trapping [26]. This enables local public health managers to predict periods of increased dengue transmission based on entomological indicators estimated through adult mosquito sampling [26, 28]. Further investigations have revealed that dengue transmission within the city is highly structured [47]. Analyzing a time series from 2017 to 2022, we observed certain areas with a higher likelihood of dengue transmission [48]. Therefore, incorporating VECTRACK sensor into an ongoing surveillance system, such as the one developed in Foz do Iguaçu, could significantly enhance its impact. In the routine entomological surveillance of Foz do Iguaçu, several hundreds of mosquitoes are sampled monthly. Assuming the mass trapping conducted by the city consists of collections in the urban environment, the mosquito diversity is low, i.e. only three species are sampled: Ae. aegypti, Ae. albopictus and Cx. quinquefasciatus. The VECTRACK sensors would be particularly useful in settings where large quantities of specimens are collected, where individual insect identification by local entomologists might be unfeasible. Installing these sensors in more vulnerable areas for dengue transmission would enable local public health managers not only to identify mosquito species and sex but also to integrate them into the early warning system algorithm [49]. This would provide alerts and prompt further interventions in areas with, for example, a high concentration of female Ae. aegypti mosquitoes and nearby human populations [13, 50,51,52,53].

Some reports have shown effective automatic classification of mosquito genera [54], sex [41] or both [36], but only a few have been able to provide point-of-capture classification. The VECTRACK sensor was used in two provinces of Barcelona, Spain, after training and developing a machine learning model using the flight of laboratory-reared Ae. albopictus and Cx. pipiens. A balanced accuracy of 95.5 and 88.8% in discriminating target mosquitoes and classifying genus/sex was observed, respectively [39]. In a recent report using semi-field conditions, the ability of an optoacoustic smart trap in classifying Ae. aegypti, Cx. quinquefasciatus and Anopheles stephensi with accuracies > 90% was shown [55]. We conducted field capturing of mixed native populations of both Ae. aegypti and Cx. quinquefasciatus in two dengue-endemic cities and observed 99.87% accuracies for classifying target species. Furthermore, the VECTRACK sensor discriminated Aedes and Culex with an accuracy of 93.72%. We also showed > 92% accuracy in discriminating mosquito sex in each of the target genera.

One remarkable outcome of our field survey in Brazil is that we used the models developed in Barcelona for Ae. albopictus and Cx. pipiens [36]. At least two conclusions can be reached regarding this result. First, the machine learning-based model for Ae. albopictus worked well in classifying Ae. aegypti mosquitoes from two Brazilian populations. The model developed and trained with Spanish and Portuguese Ae. albopictus populations worked surprisingly well for Brazilian populations of Ae. aegypti. The second remark regards classifying Cx. quinquefasciatus using the model developed and trained using Spanish populations of Cx. pipiens [36]. Culex pipiens complex comprises Culex pipiens pipiens, which exists in two forms, pipiens and molestus, along with Culex pipiens pallens, Cx. quinquefasciatus, Cx. australicus and Cx. globocoxitus. While some members of the complex have restricted geographic distributions, Cx. pipiens pipiens and Cx. quinquefasciatus are widely distributed across urban and suburban temperate and tropical regions worldwide [56]. Our findings suggest the original models developed and trained for European native mosquitoes of the Aedes and Culex genera could be extrapolated to other genetically related mosquito species from different countries. In that scenario, a further next step would involve developing specific models for Ae. aegypti and Cx. quinquefasciatus to test whether an expected increase in model accuracy in classifying target species is achieved.

The lowest correlation between manual counting and the VECTRACK sensor was observed for Cx. quinquefasciatus males. If the coefficient of determination between manual records and the VECTRACK sensor was > 0.98 for Ae. aegypti males and females, as well as for Cx. quinquefasciatus females, an R2 of 0.889 was estimated for Cx. quinquefasciatus males. One hypothesis to explain this could be related to mosquito size. It is well known in vector biology that the wing size is a proxy of mosquito size. Furthermore, the size of an adult mosquito is influenced by the quality of the breeding site in which mosquitoes have developed. In highly competitive low-resource breeding sites, mosquito size will likely be smaller than if mosquitoes are reared in a low-competitive and high-resource environment. Regarding Cx. quinquefasciatus, this species often uses large water bodies full of organic material for egg-laying and by corollary larval rearing [4, 57, 58]. Therefore, considering larval nutrients might be abundant in those breeding sites, the size of adult Cx. quinquefasciatus males that were trapped could have impacted our estimates.

The study had further limitations that should be noted. One is that we did not estimate mosquito size by measuring wing length, a common proxy adopted in medical entomology, and therefore could not test whether the size of Cx. quinquefasciatus males could negatively impact the coefficient of determination between manual counting and the VECTRACK sensor [59]. Additionally, our field sampling lasted almost 1 year, and due to abiotic factors related to seasonality, a period of 2 years of monitoring would be recommended. Nevertheless, herein we show the VECTRACK sensor is able to operate and provide robust and reasonably good classification accuracy results (93.72%) for the target Aedes and Culex mosquito species over the sampling conducted in two dengue-endemic cities of Brazil. We recommend additional tests should be done under more realistic field scenarios such as urban landscapes and areas with higher mosquito biodiversity. Furthermore, developing and training models for other mosquito vector species would be valuable as well.

Conclusions

The use of emerging technologies to improve mosquito surveillance and vector control is on the rise lately, and their use can supplement the response to mosquito-borne diseases outbreaks. Our data show the effectiveness of an optical sensor in classifying target vs. non-target insect species with an accuracy of 99.9% and a specificity of 99.9%. Using field-gathered data, we could differentiate Aedes and Culex mosquitoes with an accuracy of 93.7% and precision of 96.4%. Finally, we also determined sex differentiation for the two genera of interest: Aedes and Culex. The determination of sex in Cx. quinquefasciatus mosquitoes achieved an accuracy of 95%, whereas separating male and female Ae. aegypti was done with a 92.1% success rate. Remarkably, similar results were obtained in two different Brazilian cities, Rio de Janeiro and Brasilia, suggesting high reliability of our findings. One important feature to highlight is potential extrapolation of the original developed and trained models. The model developed and trained for European colonies of Ae. albopictus and Cx. pipiens presented high accuracy for Brazilian populations of Ae. aegypti and Cx. quinquefasciatus, respectively. Our findings show this optical sensor can be coupled with conventional traps and later integrated into mosquito surveillance methods to generate accurate automatic real-time monitoring of medically relevant mosquito species.

Availability of data and materials

All data supporting the findings of the study are available within the article.

Abbreviations

DENV:

Dengue virus

ZIKV:

Zika virus

CHIKV:

Chikungunya virus

OROV:

Oropouche virus

WNV:

West Nile virus

BG:

Biogents

VBD:

Vector-borne diseases

MAYV:

Mayaro virus

ULV:

Ultra-low volume

LED:

Light-emitting diodes

ADS:

Analog to digital converter

R 2 :

Coefficient of determination

TP:

True positive

TN:

True negative

FP:

False positive

FN:

False negative

References

  1. Brady OJ, Hay SI. The global expansion of dengue: how Aedes aegypti mosquitoes enabled the first pandemic arbovirus. Annu Rev Entomol. 2020;65:191–208.

    Article  CAS  PubMed  Google Scholar 

  2. Weaver SC, Reisen WK. Present and future arboviral threats. Antiviral Res. 2010;85:328–45.

    Article  CAS  PubMed  Google Scholar 

  3. Kraemer MUG, Sinka ME, Duda KA, Mylne AQN, Shearer FM, Barker CM, et al. The global distribution of the arbovirus vectors Aedes aegypti and Ae. albopictus. Elife 2015;4:1–18.

  4. David MR, Ribeiro GS, de Freitas RM. Bionomics of Culex quinquefasciatus within urban areas of Rio de Janeiro. Southeastern Brazil Rev Saude Publica. 2012;46:858–65.

    Article  PubMed  Google Scholar 

  5. David MR, Dantas ES, Maciel-de-Freitas R, Codeço CT, Prast AE, Lourenço-de-Oliveira R. Influence of larval habitat environmental characteristics on Culicidae immature abundance and body size of adult Aedes aegypti. Front Ecol Evol. 2021;9:1–12.

    Article  Google Scholar 

  6. Abílio AP, Abudasse G, Kampango A, Candrinho B, Sitoi S, Luciano J, et al. Distribution and breeding sites of Aedes aegypti and Aedes albopictus in 32 urban/peri-urban districts of Mozambique: implication for assessing the risk of arbovirus outbreaks. PLoS Negl Trop Dis. 2018;12:e0006692.

    Article  PubMed  PubMed Central  Google Scholar 

  7. Maciel-de-Freitas R, Brocki-Neto R, Gonçalves JM, Codeço CT, Lourenço-de-Oliveira R. Movement of dengue vectors between the human modified environment and an urban forest in Rio de Janeiro. J Med Entomol. 2006;43:1112–20.

    Article  PubMed  Google Scholar 

  8. Edman JD, Strickman D, Kittayapong P, Scott TW. Female Aedes aegypti (Diptera: Culicidae) in Thailand rarely feed on sugar. J Med Entomol. 1992;29:1035–8.

    Article  CAS  PubMed  Google Scholar 

  9. Baldacchino F, Caputo B, Chandre F, Drago A, della Torre A, Montarsi F, et al. Control methods against invasive Aedes mosquitoes in Europe: a review. Pest Manag Sci. 2015;71:1471–85.

    Article  CAS  PubMed  Google Scholar 

  10. Pereira dos Santos T, Roiz D, Santos de Abreu FV, Luz SLB, Santalucia M, Jiolle D, et al. Potential of Aedes albopictus as a bridge vector for enzootic pathogens at the urban-forest interface in Brazil. Emerg Microbes Infect. 2018;7:1–8.

  11. Lounibos LP, Kramer LD. Invasiveness of Aedes aegypti and Aedes albopictus and vectorial capacity for chikungunya virus. J Infect Dis. 2016;214:S453–8.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL, et al. The global distribution and burden of dengue. Nature. 2013;496:504–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Louis VR, Phalkey R, Horstick O, Ratanawong P, Wilder-Smith A, Tozan Y, et al. Modeling tools for dengue risk mapping - a systematic review. Int J Health Geogr. 2014;13:50.

    Article  PubMed  PubMed Central  Google Scholar 

  14. Lorenz C, Almeida F, Almeida-Lopes F, Louise C, Pereira SN, Petersen V, et al. Geometric morphometrics in mosquitoes: What has been measured? Infect Genet Evol. 2017;54:205–15.

    Article  PubMed  Google Scholar 

  15. Esposito DLA, da Fonseca BAL. Will Mayaro virus be responsible for the next outbreak of an arthropod-borne virus in Brazil? Brazilian J Infect Dis. 2017;21:540–4.

    Article  Google Scholar 

  16. Vasconcelos PFC, Calisher CH. Emergence of Human Arboviral Diseases in the Americas, 2000–2016. Vector-Borne Zoonotic Dis. 2016;16:295–301.

    Article  PubMed  Google Scholar 

  17. Jourdain F, Samy AM, Hamidi A, Bouattour A, Alten B, Faraj C, et al. Towards harmonisation of entomological surveillance in the Mediterranean area. PLoS Negl Trop Dis. 2019;13:e0007314.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Roiz D, Wilson AL, Scott TW, Fonseca DM, Jourdain F, Müller P, et al. Integrated Aedes management for the control of Aedes-borne diseases. PLoS Negl Trop Dis. 2018;12:e0006845.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Moyes CL, Vontas J, Martins AJ, Ng LC, Koou SY, Dusfour I, et al. Contemporary status of insecticide resistance in the major Aedes vectors of arboviruses infecting humans. Sinnis P, editor. PLoS Negl Trop Dis. 2017;11:e0005625.

  20. Chala B, Hamde F. Emerging and re-emerging vector-borne infectious diseases and the challenges for control: A review. Front Public Heal. 2021;9:1–10.

    Google Scholar 

  21. Maciel-de-Freitas R, Avendanho FC, Santos R, Sylvestre G, Araújo SC, Lima JBP, et al. Undesirable consequences of insecticide resistance following Aedes aegypti control activities due to a dengue outbreak. PLoS ONE. 2014;9:e92424.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Ellwanger JH, Kulmann-Leal B, Kaminski VL, Valverde-Villegas JM, Veiga ABG, Spilki FR, et al. Beyond diversity loss and climate change: Impacts of Amazon deforestation on infectious diseases and public health. An Acad Bras Cienc. 2020;92:1–33.

    Article  Google Scholar 

  23. Codeço CT, Lima AWS, Araújo SC, Lima JBP, Maciel-de-Freitas R, Honório NA, et al. Surveillance of Aedes aegypti: Comparison of house index with four alternative traps. PLoS Negl Trop Dis. 2015;9:e0003475.

    Article  PubMed  PubMed Central  Google Scholar 

  24. da Cruz Ferreira DA, Degener CM, de Almeida M-T, Bendati MM, Fetzer LO, Teixeira CP, et al. Meteorological variables and mosquito monitoring are good predictors for infestation trends of Aedes aegypti, the vector of dengue, chikungunya and Zika. Parasit Vectors. 2017;10:78.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Cromwell EA, Stoddard ST, Barker CM, Van Rie A, Messer WB, Meshnick SR, et al. The relationship between entomological indicators of Aedes aegypti abundance and dengue virus infection. PLoS Negl Trop Dis. 2017;11:e0005429.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Leandro AS, de Castro WAC, Lopes RD, Delai RM, Villela DAM, De-Freitas RM. Citywide integrated Aedes aegypti mosquito surveillance as early warning system for arbovirus transmission Brazil. Emerg Infect Dis. 2022;28:701–6.

    Article  PubMed  Google Scholar 

  27. de Leandro A, et al. The adoption of the One Health approach to improve surveillance of venomous animal injury, vector-borne and zoonotic diseases in Foz do Iguaçu Brazil. PLoS Negl Trop Dis. 2021;15:0009109.

    Article  Google Scholar 

  28. de Leandro A, et al. Optimising the surveillance of Aedes aegypti in Brazil by selecting smaller representative areas within an endemic city. Trop Med Int Heal. 2024;29:1–10.

    Google Scholar 

  29. Lau S-M, Vythilingam I, Doss JI, Sekaran SD, Chua TH, Wan Sulaiman WY, et al. Surveillance of adult Aedes mosquitoes in Selangor. Malaysia Trop Med Int Heal. 2015;20:1271–80.

    Article  CAS  Google Scholar 

  30. Sivagnaname N, Gunasekaran K. Need for an efficient adult trap for the surveillance of dengue vectors. Indian J Med Res. 2012;136:739–49.

    CAS  PubMed  PubMed Central  Google Scholar 

  31. Villela DAM, de Garcia G. Novel inference models for estimation of abundance, survivorship and recruitment in mosquito populations using mark-release-recapture data. PLoS Negl Trop Dis. 2017;11:0005682.

    Article  Google Scholar 

  32. Chadee DD, Ritchie SA. Efficacy of sticky and standard ovitraps for Aedes aegypti in Trinidad. West Indies J Vector Ecol. 2010;35:395–400.

    Article  PubMed  Google Scholar 

  33. Ritchie SA, Long S, Smith G, Pyke A, Knox TB. Entomological investigations in a focus of dengue transmission in Cairns, Queensland, Australia, by using the sticky ovitraps. J Med Entomol. 2004;41:1–4.

    Article  PubMed  Google Scholar 

  34. Service MW. Mosquito Ecology. 2nd ed. Dordrecht: Springer, Netherlands; 1993.

    Book  Google Scholar 

  35. Maciel-de-Freitas R, Peres RC, Alves F, Brandolini MB. Mosquito traps designed to capture Aedes aegypti (Diptera: Culicidae) females: preliminary comparison of Adultrap, MosquiTRAP and backpack aspirator efficiency in a dengue-endemic area of Brazil. Mem Inst Oswaldo Cruz. 2008;103:602–5.

    Article  PubMed  Google Scholar 

  36. González-Pérez MI, Faulhaber B, Williams M, Brosa J, Aranda C, Pujol N, et al. A novel optical sensor system for the automatic classification of mosquitoes by genus and sex with high levels of accuracy. Parasit Vectors. 2022;15:190.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Krockel U, Rose A, Eiras AE, Geier M. New tools for surveillance of adult yellow fever mosquitoes: comparison of trap catches with human landing rates in an urban environment. J Am Mosq Control Assoc. 2006;22:229–38.

    Article  PubMed  Google Scholar 

  38. Maciel-de-Freitas R, Eiras ÁE, Lourenço-de-Oliveira R. Field evaluation of effectiveness of the BG-Sentinel, a new trap for capturing adult Aedes aegypti (Diptera: Culicidae). Mem Inst Oswaldo Cruz. 2006;101:321–5.

    Article  PubMed  Google Scholar 

  39. González-Pérez MI, Faulhaber B, Aranda C, Williams M, Villalonga P, Silva M, et al. Field evaluation of an automated mosquito surveillance system which classifies Aedes and Culex mosquitoes by genus and sex. Parasit Vectors. 2024;17:97.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Consoli RAGB, Lourenço-de-Oliveira R. Principais mosquitos de importância sanitária no Brasil. Fiocruz, editor. Rio de Janeiro; 1994.

  41. Genoud AP, Basistyy R, Williams GM, Thomas BP. Optical remote sensing for monitoring flying mosquitoes, gender identification and discussion on species identification. Appl Phys B. 2018;124:57.

    Article  Google Scholar 

  42. Genoud AP, Gao Y, Williams GM, Thomas BP. A comparison of supervised machine learning algorithms for mosquito identification from backscattered optical signals. Ecol Inform. 2020;58:101090.

    Article  Google Scholar 

  43. Genoud AP, Gao Y, Williams GM, Thomas BP. Identification of gravid mosquitoes from changes in spectral and polarimetric backscatter cross sections. J Biophotonics. 2019;12:e201900123.

    Article  PubMed  PubMed Central  Google Scholar 

  44. Barcellos C, Matos V, Lana RM, Lowe R. Climate change, thermal anomalies, and the recent progression of dengue in Brazil. Sci Rep. 2024;14:5948.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Lowe R, Barcellos C, Brasil P, Cruz O, Honório N, Kuper H, et al. The Zika virus epidemic in Brazil: from discovery to future implications. Int J Environ Res Public Health. 2018;15:96.

    Article  PubMed  PubMed Central  Google Scholar 

  46. Leandro A de S, Lopes RD, Amaral Martins C, Delai RM, Villela DAM, Maciel-de-Freitas R. Entomovirological surveillance followed by serological active survey of symptomatic individuals is helpful to identify hotspots of early arbovirus transmission. Front Public Heal. 2022;10:1024187.

  47. Leandro AS, Chiba de Castro WA, Garey MV, Maciel-de-Freitas R. Spatial analysis of dengue transmission in an endemic city in Brazil reveals high spatial structuring on local dengue transmission dynamics. Sci Rep. 2024;14:8930.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  48. de Souza A, Leandro MJC, Ayala RD, Lopes CA, Martins RMF, Villela DAM. Entomo-virological Aedes aegypti surveillance applied for prediction of dengue transmission: a spatio-temporal modeling study. Pathogens. 2022;12:4.

    Article  Google Scholar 

  49. Coelho FC, Codeço CT. Precision epidemiology of arboviral diseases. J Public Heal Emerg. 2019;3:1–1.

    Article  Google Scholar 

  50. Olliaro P, Fouque F, Kroeger A, Bowman L, Velayudhan R, Santelli AC, et al. Improved tools and strategies for the prevention and control of arboviral diseases: a research-to-policy forum. PLoS Negl Trop Dis. 2018;12:e0005967.

    Article  PubMed  PubMed Central  Google Scholar 

  51. Hussain-Alkhateeb L, Rivera Ramírez T, Kroeger A, Gozzer E, Runge-Ranzinger S. Early warning systems (EWSs) for chikungunya, dengue, malaria, yellow fever, and Zika outbreaks: What is the evidence? A scoping review. PLoS Negl Trop Dis. 2021;15:e0009686.

    Article  PubMed  PubMed Central  Google Scholar 

  52. Lee K-S, Lai Y-L, Lo S, Barkham T, Aw P, Ooi P-L, et al. Dengue virus surveillance for early warning. Singapore Emerg Infect Dis. 2010;16:847–9.

    Article  PubMed  Google Scholar 

  53. Brady OJ, Johansson MA, Guerra CA, Bhatt S, Golding N, Pigott DM, et al. Modelling adult Aedes aegypti and Aedes albopictus survival at different temperatures in laboratory and field settings. Parasit Vectors. 2013;6:351.

    Article  PubMed  PubMed Central  Google Scholar 

  54. Potamitis I, Rigakis I. Measuring the fundamental frequency and the harmonic properties of the wingbeat of a large number of mosquitoes in flight using 2D optoacoustic sensors. Appl Acoust. 2016;109:54–60.

    Article  Google Scholar 

  55. Johnson BJ, Weber M, Al-Amin HM, Geier M, Devine GJ. Automated differentiation of mixed populations of free-flying female mosquitoes under semi-field conditions. Sci Rep. 2024;14:3494.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  56. Farajollahi A, Fonseca DM, Kramer LD, Marm KA. “Bird biting” mosquitoes and human disease: a review of the role of Culex pipiens complex mosquitoes in epidemiology. Infect Genet Evol. 2011;11:1577–85.

    Article  PubMed  PubMed Central  Google Scholar 

  57. Chevillon C. Population genetics of insecticide resistance in the mosquito Culex pipiens. Biol J Linn Soc. 1999;68:147–57.

    Article  Google Scholar 

  58. Vilibic-Cavlek T, Savic V, Petrovic T, Toplak I, Barbic L, Petric D, et al. Emerging trends in the epidemiology of west nile and usutu virus infections in Southern Europe. Front Vet Sci. 2019;6:437.

    Article  PubMed  PubMed Central  Google Scholar 

  59. Harbach RE, Knight KL. Taxonomist’s glossary of mosquito anatomy. Medford: Plexus Publications Co; 1980. p. 1–54.

    Google Scholar 

Download references

Acknowledgements

We thank those responsible in the Astral Saúde Ambiental and the National Health Surveillance Department for allowing the BG-Mosquitaire + Vectrack sensor installation at Rio de Janeiro and Brasilia, respectively.

Funding

This research was supported by the project E4WARNING. This project has received funding from the European Union’s Horizon Europe programme (HORIZON Research and Innovation Actions) under Grant Agreement 101086640 and the Fundação Carlos Chagas Filho de Amparo à Pesquisa no Estado do Rio de Janeiro, Grant E-26/201.075/2022.

Author information

Authors and Affiliations

Authors

Contributions

Conceptualization F.C.B.F.P.N, R.C.M. and R.M.F.; methodology F.C.B.F.P.N. and R.C.M.; formal analysis A.S.L. and R.M.F.; investigation F.C.B.F.P.N, R.C.M., A.S.L., and R.M.F.; resources, F.C.B.F.P.N and R.M.F.; writing and original draft preparation R.M.F.; writing—review and editing F.C.B.F.P.N, R.C.M., A.S.L. and R.M.F.; supervision A.S.L. and R.M.F. All authors have read and agreed to the published version of the manuscript.

Corresponding author

Correspondence to Rafael Maciel-de-Freitas.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Njaime, F.C.B.F.P., Máspero, R.C., Leandro, A.d. et al. Automated classification of mixed populations of Aedes aegypti and Culex quinquefasciatus mosquitoes under field conditions. Parasites Vectors 17, 399 (2024). https://doi.org/10.1186/s13071-024-06417-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13071-024-06417-z

Keywords