Progress 09/01/22 to 12/31/24

Outputs
Target Audience:The National National Association of County and City Health Officials (NACCHO) estimates there are over 2000 mosquito control organizations in the US. Mosquito control, typically conducted by county and municipal departments of health, is particularly challenging in the context of rural communities where they are often responsible for managing the largest areas while having the fewest resources. As a result of limited surveillance resources and therefore limited control capabilities, rural and economically disadvantaged regions are less likely to be capable of preventing, mitigating, and responding to mosquito-borne disease outbreaks. The successful development of an automated counting trap for mosquito surveillance will benefit not only local public health stakeholders responsible for managing mosquito control programs but also reduce mosquito-borne disease risk and nuisance mosquito impact on the community. During this period, we have worked with MCOs to provide design feedback and usability insights to help us provide a product well tailored to their needs. Changes/Problems:There were a few unanticipated problems that delayed the product development schedule and pushed the timeline for additional image data gathering and algorithm development beyond the scheduled end date of the grant period. The first of which was the pivot to an alternatesingle board computer due to unanticipatedsupply-chain issues, thereby delayingthe development schedule. Additionally, cascading issues with communication and overheating delayed image gathering with partners. Collecting sufficient image data is a prerequisite to algorithm development, and so computer vision development was delayed. Despite these delays, with additional time granted with the NCE, additional data allowed for significant progress on algorithm development. High fidelity object detection and species classification was validated, paving the way for the final development stages of Scout prior to generalized usage in target regions. What opportunities for training and professional development has the project provided?The work accomplished under this grant has created ample professional development opportunities for the individuals on the project, spanning the fields of entomology and mechanical, electrical, software, and product engineering. Entomologists at both the internship and professional level have contributed to this work, gaining valuable insectary, field testing and product development experience. Vectech supported two summer mechanical engineering internships during which two undergraduate mechanical engineering students tackled technical issues relevant to this work, including a project on cooling of electronics as well as motor design and failure testing. With regards to professional technical mentorship, senior staff at Vectech have provided 1 to 1 mentorship to the entry and intermediate members of the team. Additionally, we have worked with Root3 Lab engineering design firm to leverage their expertise in product design and electrical engineering, providing mentorship and consultation in these areas. How have the results been disseminated to communities of interest?A description of the technology and current status was communicated to interested parties via a technical presentation at the American Mosquito Control Association conference in March 2024. Additionally, we have worked collaboratively with mosquito control organizations (MCOs) throughout the design and optimization of the system. In addition to conversations at conferences and symposiums, we have had meetings with individuals at MCOs to describe the technology. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Impact statement: A novel remote sensing system for identification and counting of mosquitoes as they enter a trap hasbeen developed. The system is able to remotely transmit abundance information to community public health agencies and provide actionable biosurveillance data, overcoming resource limitations of traditional surveillance. Objective 1: Build functional prototypes through integration of the phase I sensing system into a mosquito trap built to the identified specifications for a minimum viable product. Functional prototypes were built and verified to meet the product requirements. The system has been tested, and usability feedback and results of semi-field and field testing analyzed to improve the design. The following is a list of activities performed during the full grant timeframe under this objective. Built functional prototypes that achieve the optical, illumination, airflow, and durability design specifications. Designed the system to be sufficiently attractive to mosquitos and to effectively constrain them for high-quality imaging Changed color of external parts of custom trap to leverage visual cues for attracting certain species Leveraged heat of processor board to attract mosquitoes and design airflow to move warm air towards entrance of the trap Tested attractivity in comparison to the BG-pro trap in semi-field and field settings Optimized entrance funnel geometry to reduce power consumption of the fan while achieving sufficient airflow at the inlet of the trap Designed the system for robustness in the field and requirement for mean time between failure (MTBF) Selected components for the intended number of cycles of the mesh chamber subsystem Performed failure analysis and implemented design changes, iterating components and material type and performing verification through cycle testing Stepper motor selected for rotating platform as it would minimize the wear on the custom mesh chamber parts Carriedout necessary hardware design improvements ?Implemented peripheral board and ribbon cable design that centralizes the point the peripherals connect, streamlining and reducing the complexity of the wiring and room for error, thereby expediting the manufacturing of the system. Design power board that centralized the power input and supply for the electrical components within the system Improved overall structure of electronics mount and included wire management within the system Implementation of active cooling within the system via an additional cooling fan that maintains the electronicsoperating temperature Selection and verification of a lower power, lower volume trapping fan Design for weatherproofing and meeting IP design requirements including changes to the structure, addition of gaskets White cone attached to bottom side of mesh chamber with a twist mechanism for easy maintenance and replacement. Structural updates were included in this iteration. Firmware improvements Selected appropriate processors Update firmware formore consistent and reliable iot imaging Selected camera board that meets specifications and was validated to be compatible with the sbc Capture settings identified through experimentation and expert review by senior entomologist and computer vision engineers Creation of the web-dashboard interface that included the following features Allows user to connect device to wifi via the web dashboard Displays images captured with the system Displays mechanical and upload errors that occur Allows the user to set and change capture schedule and frequency Objective 2: Collect representative image data to iterate, train, and deploy computer vision models to count and identify mosquitoes in the functional prototype. We've described the image data gathering activities performed during this reporting period below, as well as algorithm development activities that were achieved to enable species classification and count data. A robust image dataset provides the foundation for development of high-accuracy algorithms that extract species and count information. One-hundred lab-reared Aedes aegypti and 50 Aedes albopictus have been imaged within Scout, providing image data for algorithm development and training. Wild-caught image data was collected through field testing in Maryland during August 2024. Approximately 50 mosquitoes were imaged during this time, as well as by-catch (non-mosquito insects). A dataset consisting of colony strains of the three distinct species, Culex quinquefasciatus (n=513), Anopheles gambiae (n=406), and Aedes aegypti (n=1665), and 507 images of bycatch image data (non-mosquitoes). A two stage computer vision pipeline was developed consisting of Yolov8 object detection model that detects mosquitos and counts the number of mosquitos present in an image. The second stage of the algorithm is an Xception model enabling species classification. Enhanced mosquito detection capability in the field environment was achieved by introducing 'bycatch' detection, for distinguishing non-mosquitoes also captured in the trap. The object detection model was able to distinguish between bycatch and mosquito with a 100% accuracy. The model achieved a recall of 85% on mosquito dataset, correctly counting 85% of the 881 mosquitoes it was evaluated on, and a precision of 93%, in which 93% of its mosquito predictions were correct. It also achieved a recall of 95% and a precision of 92% on bycatch. The classification model was successfully able to classify between the three mosquito species mentioned above with over 99% accuracy. Objective 3: Validate the usability, functionality, and performance of prototypes and the deployed computer vision system through a three-month field test with partners. Testing to assess the usability, functionality and performance of interactive prototypes were performed over the course of the spring and summer months of 2024. Rather than a single, continuous three-month field test, these tests were broken down into iterative tests that informed subsequent designs. Assembled the trap hardware to facilitate the usability and functionality testing 3 alpha prototypes were designed and assembled for semi-field testing as well as for design and usability testing with partners and to validate the concept and functionality of the system Based on feedback from the alpha prototypes, 5 beta prototypes were designed and assembled for functionality and field testing. These traps were built to be weatherproof (meeting IP requirements) and robust Functionality and performance testing were completed at both the alpha and beta development stages Semifield testing in partnership with Anastasia Mosquito Control District in April 2024 resulted in capture efficiency of the system in comparison to the CDC light trap and BG-Pro traps. Iterative in-house reliability and functionality testing to assess adherence to performance as outlined in the product requirements Field testing in Baltimore, MD in August 2024 to assess weatherproofing and field-performance requirements Usability testing was performed as both the alpha and beta development stages Usability sessions with 10 stakeholders during the 2024 American Mosquito Control conference, including MCO and research officials in CA, UT, NV, FL, KS, and NC. Usability testing with Anastasia Mosquito in April 2024 Usability and concept interviews with stakeholders performed virtually for acquiring feedback Internal usability sessions with in-house entomologists using the trap in field settings

Publications

Progress 09/01/23 to 08/31/24

Outputs
Target Audience:The National National Association of County and City Health Officials (NACCHO) estimates there are over 2000 mosquito control organizations in the US. Mosquito control, typically conducted by county and municipal departments of health, is particularly challenging in the context of rural communities where they are often responsible for managing the largest areas while having the fewest resources. As a result of limited surveillance resources and therefore limited control capabilities, rural and economically disadvantaged regions are less likely to be capable of preventing, mitigating, and responding to mosquito-borne disease outbreaks. The successful development of an automated counting trap for mosquito surveillance will benefit not only local public health stakeholders responsible for managing mosquito control programs but also reduce mosquito-borne disease risk and nuisance mosquito impact on the community. During this period, we have worked with MCOs to provide design feedback and usability insights to help us provide a product well tailored to their needs. Changes/Problems:There were a few unanticipated problems that delayed the product development schedule and pushed the timeline for additional image data gathering and algorithm development beyond the scheduled end date of the grant period. The first of which was the pivot to the BeagleBoneAI-64 single board computer that is used in the beta prototype. Because of the risk of supply-chain issues with Chinese-produced products as well as prior selected options going out of stock, we were forced to pivot to a new architecture, which delayed the development schedule. Additionally, cascading issues with communication and overheating delayed image gathering with partners. Collecting sufficient image data is a prerequisite to algorithm development, and so computer vision development was delayed. Despite these delays, with additional time granted with the NCE, additional data will be acquired and with this data the artificial intelligence algorithms for localization and detection of the mosquitoes in the system will be developed. What opportunities for training and professional development has the project provided?The work accomplished under this grant has created ample professional development opportunities for the individuals on the project, spanning the fields of entomology and mechanical, electrical, software, and product engineering. Entomologists at both the internship and professional level have contributed to this work, gaining valuable insectary, field testing and product development experience. Vectech supported two summer mechanical engineering internships during which two undergraduate mechanical engineering students tackled technical issues relevant to this work, including a project on cooling of electronics as well as motor design and failure testing. With regards to professional technical mentorship, senior staff at Vectech have provided 1 to 1 mentorship to the entry and intermediate members of the team. Additionally, we have worked with Root3 Lab engineering design firm to leverage their expertise in product design and electrical engineering, providing mentorship and consultation in these areas. How have the results been disseminated to communities of interest?A description of the technology and current status was communicated to interested parties via a technical presentation at the American Mosquito Control Association conference in March 2024. Additionally, we have worked collaboratively with mosquito control organizations (MCOs) throughout the design and optimization of the system. In addition to converstations at conferences and symposiums, we have had meetings with individuals at MCOs to describe the technology. What do you plan to do during the next reporting period to accomplish the goals?Work during the NCE period will be focused on image data generation and algorithm development. We will focus on imaging a greater number of mosquitoes with the trap across a greater range of species. The images that are collected will need to be labeled manually with a mask that describes the location of the mosquito on the mesh as well as tied to the classification data - i.e. the species information - associated with that mosquito. Additionally, the data handling and processing pipelines, object detection algorithm, and classification algorithm will be developed and validated.

Impacts
What was accomplished under these goals? Objective 1: Build functional prototypes through integration of the phase I sensing system into a mosquito trap built to the identified specifications for a minimum viable product. During year two, we entered the beta prototype development stage, taking a prototype that is functional and iterating the design to achieve and verify product requirements. The system has been built according to product requirements and updated and improved based on usability feedback and results of semi-field and field testing. The following is a list of activities performed during this timeframe under this objective. Designed the system to be sufficiently attractive to mosquitos and to effectively constrain them for high-quality imaging Changed color of external parts of custom trap to leverage visual cues for attracting certain species Leveraged heat of processor board to attract mosquitoes and design airflow to move warm air towards entrance of the trap Tested attractivity in comparison to the BG-pro trap in semi-field and field settings Optimized entrance funnel geometry to reduce power consumption of the fan while achieving sufficient airflow at the inlet of the trap Design the system for robustness in the field and requirement for mean time between failure (MTBF) Selected components for the intended number of cycles of the mesh chamber subsystem Performed failure analysis and implemented design changes, iterating components and material type and performing verification through cycle testing Stepper motor selected for rotating platform as it would minimize the wear on the custom mesh chamber parts Carry out necessary hardware design improvements Replace BG-pro components that had been formerly incorporated to facilitate manufacturing at scale and overcome unnecessary design constraints. Implemented peripheral board and ribbon cable design that centralizes the point the peripherals connect, streamlining and reducing the complexity of the wiring and room for error, thereby expediting the manufacturing of the system. Design power board that centralized the power input and supply for the electrical components within the system Improved overall structure of electronics mount and included wire management within the system Implementation of active cooling within the system via an additional cooling fan that maintains the Beaglebone's operating temperature Selection and verification of a lower power, lower volume trapping fan Design for weatherproofing and meeting IP design requirements including changes to the structure, addition of gaskets White cone attached to bottom side of mesh chamber with a twist mechanism for easy maintenance and replacement. Structural updates were included in this iteration. Firmware improvements Selected the BBai64 and esp that together would supply sufficient processing power and be suitable in terms of supply chain Update firmware for BBai64 and esp processors for more consistent and reliable iot imaging Selected camera board that meets specifications and was validated to be compatible with the BBai64 Capture settings identified through experimentation and expert review by senior entomologist and computer vision engineers Creation of the web-dashboard interface that includesthe following features Allows user to connect device to wifi via the web dashboard Displays images captured with the system Displays mechanical and upload errors that occur Allows the user to set and change capture schedule and frequency Objective 2: Collect representative image data to iterate, train, and deploy computer vision models to count and identify mosquitoes in the functional prototype. We've described the image data gathering activities performed during this reporting period below. The image data collection activities in this period have resulted in approximately 200 images that can be used for training of the computer vision algorithms. Additional image data collection, labeling, and algorithm development will be completed during the NCE portion of this project. One-hundred lab-reared Aedes aegypti and 50 Aedes albopictus have been imaged within Scout, providing image data for algorithm development and training. These lab-reared mosquito specimens were reared in the Vectech facilities and varied in age and condition across the population. These specimens were imaged live and were allowed a natural approach to the trap entrance to ensure as close an alignment as possible to the field condition. Wild-caught image data was collected through field testing in Maryland during August 2024. Approximately 50 mosquitoes were imaged during this time, as well as by-catch that entered the trap during that period. This data can be used for object detection algorithm development and in certain instances can be tagged with species labels for the species classification algorithm development. Objective 3: Validate the usability, functionality, and performance of prototypes and the deployed computer vision system through a three-month field test with partners. Testing to assess the usability, functionality and performance of interactive prototypes were performed over the course of the spring and summer months of 2024. Rather than a single, continuous three-month field test, these tests were broken down into iterative tests that informed subsequent designs. Assembled the trap hardware to facilitate the usability and functionality testing 3 alpha prototypes were designed and assembled for semi-field testing as well as for design and usability testing with partners and to validate the concept and functionality of the system Based on feedback from the alpha prototypes, 5 beta prototypes were designed and assembled for functionality and field testing. These traps were built to be weatherproof (meeting IP requirements) and robust Functionality and performance testing were completed at both the alpha and beta development stages Semifield testing in partnership with Anastasia Mosquito Control District in April 2024 allowed us to measure capture efficiency of the system in comparison to the CDC light trap and BG-Pro traps. Iterative in-house reliability and functionality testing to assess adherence to performance as outlined in the product requirements Field testing in Baltimore, MD in August 2024 to assess weatherproofing and field-performance requirements Usability testing was performed as both the alpha and beta development stages Usability sessions with 10 stakeholders during the 2024 American Mosquito Control conference, including MCO and research officials in CA, UT, NV, FL, KS, and NC. Usability testing with Anastasia Mosquito in April 2024 Usability and concept interviews with stakeholders performed virtually for acquiring feedback Internal usability sessions with in-house entomologists using the trap in field settings

Publications

Progress 09/01/22 to 08/31/23

Outputs
Target Audience:The National National Association of County and City Health Officials (NACCHO) estimates there are over 2000 mosquito control organizations in the US. Mosquito control, typically conducted by county and municipal departments of health, is particularly challenging in the context of rural communities where they are often responsible for managing the largest areas while having the fewest resources. As a result of limited surveillance resources and therefore limited control capabilities, rural and economically disadvantaged regions are less likely to be capable of preventing, mitigating, and responding to mosquito-borne disease outbreaks. The successful development of an automated counting trap for mosquito surveillance will benefit not only local public health stakeholders responsible for managing mosquito control programs but also reduce mosquito-borne disease risk and nuisance mosquito impact on the community. During this period, we have worked with MCOs to inform the design specifications, asking for feedback and guidance to design a product well-tailored to their needs. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?The development process for the SCOUT trap has provided ample professional development opportunities for the engineers through individual study. In particular, the manufacturing knowledge of the team has strengthened. Through research and review, we have developed improved processes for documentation and transfer to manufacturing. In addition, we have mentors that have provided 1-to-1 mentorship of our technical team. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?Subsequent hardware and software development is required to be ready to hand off the prototype to be used by external users. This development will take place over the next 4 months. With these prototypes ready in the spring, we will focus on image data gathering with partners. This will involve the capture and identification of wild larvae, and the staged semi-field releases into the trap. A larger data set will allow for more rigorous and tailored computer vision development. Concurrently, design for manufacture will be considered before the 15-device manufacturing run in preparation for field testing in the summer. A detailed study will be prepared before the field testing.

Impacts
What was accomplished under these goals? Objective 1: Build functional prototypes through integration of the phase I sensing system into a mosquito trap built to the identified specifications for a minimum viable product. During year one, we designed and built a functional prototype of Scout that achieves the optical, illumination, airflow, and durability design specifications. The prototype captures a white-background image with an optimized background profile and features components that improve specimen illumination. Additionally, we have developed the firmware for controlling the IoT trap, which allows for remote image capture, mesh rotation, and camera setting adjustment, among other features. Objective 2: Collect representative image data to iterate, train, and deploy computer vision models to count and identify mosquitoes in the functional prototype. We have begun collection and imaging of a small set of live wild-caught or first-generation reared mosquitoes to build a representative image dataset. We have planned partnerships for next spring that will aid us in finding a diverse set of specimens and building an extensive dataset for training. With this data, we will be able to design robust algorithms. Algorithm development is pending further growth of the initial image dataset. Objective 3: Validate the usability, functionality, and performance of prototypes and the deployed computer vision system through a three-month field test with partners. Work for objective three will take place in 2024.

Publications