Dengue is one of the most prominent mosquito-borne diseases in the world. This illness is endemic in the Philippines and is constantly a serious threat to the health of millions of Filipinos every day. The Philippines, being a tropical and humid country, holds the perfect conditions as a breeding ground for mosquitos. With the increasing population size and density of local communities and the ongoing urbanization of rural and semi-urban areas, there are more people at risk becoming infected with the disease. Alongside the health risks that dengue cause with the annual mosquito outbreaks, there are economic burdens on the country as well as mentioned by Edillo et al. (2015). Large number of dengue cases prove to be difficult for the country’s health institutions to bear, especially in rural areas that are less medically equipped and prepared to combat such illness. Many rural health centers do not have the technology and resources to handle outbreaks consistently, so in order to stave the number of cases, the local government invested in preventive measures in order to counter dengue outbreaks.
The government alongside other institutions have implemented mosquito surveillance systems across the Philippines (Undurraga et al., 2017). These systems detect dengue and mosquito outbreaks through an early warning system that is based on local reports and cases of dengue which are provided by disease reporting units such as local government units, municipal or city health centers, hospitals and private clinics. Most of the data focuses on incidents and reports coming from cases of quarantined dengue patients. This provides a high degree of certainty when it comes to confirming outbreaks, however it takes time for the medical results to be processed and relies too much on sufficient data to be collected before confirmation. Furthermore, processing blood tests for confirmation of outbreaks takes days before being delivered. Any action to deal with dengue outbreaks are delayed just by waiting for the results to be processed and sent. There is also the case for dengue episodes that are left unreported which hinder any potential early responses by government units and of accurate estimation of the spread of dengue (Undurraga et al., 2017).
With the changing climate and weather patterns, relying on seasonal changes for mosquito outbreaks to determine and prevent dengue is becoming more difficult. Rainy season in the Philippines has become less consistent, and has become much more erratic making it difficult to pinpoint mosquito breeding seasons until it is too late. More preparation will be required to quickly dispatch anti-mosquito operations and take care of dengue outbreaks early on if said outbreaks can only be determined once it is occurring using the current surveillance system. In order to effectively counter this, the focus for preventive measures have shifted from cases of dengue to the outbreaks of mosquitos. Mosquito traps and lures, were set up in many key areas. Once mosquitos are successfully lured, the traps’ contents are analyzed manually by experts. Data on the mosquitos are collected and from that it can help determine if a mosquito outbreak and potentially, when dengue is on the rise. With the data, local governments can prevent dengue outbreaks by dealing with the mosquitos first.
Current dengue and mosquito surveillance systems are able to lessen the degree of infection among the population by notifying local governments quickly. The speed at which the notifications are received are greatly limited by the amount of time it takes to analyze and confirm the data of changes in mosquito populations and reports of incidents of dengue. Observation and analysis of such data, especially that of mosquitos, commonly involves the assistance of experts in the field entomology to accurately classify and discriminate mosquitos via visual judgement and acquired knowledge on mosquitos. Manual classification takes time since not only do the insects need to be classified, the mosquitos that are successfully detected need to be further discriminated to determine the likelihood carrying dengue and potentially other mosquito-borne illnesses. Aside from being time consuming, manual work is error prone and labor intensive. If time spent analyzing data can be lessened, it can speed up the process and allow for early response times for preventive measures and for a more streamline approach for mosquito detection.
Within the last decade, the field of computer vision has grown in popularity with multiple studies leaning towards machine learning implementations. Image classification and object detection, are two of the most notable topics for these implementations and are applied in various fields of study. One such implementation is the usage of convolutional neural networks to extract high-level information from an image's raw pixel data as input data for the model. This ultimately helps the model learn and extract key features for the inference object classes. This study plans to make use of convolutional neural networks in creating and determining a general workflow for the discrimination, identification and counting of mosquitos.