Delimiting cryptic morphological variation among human malaria vector species using convolutional neural networks

Automated differentiation of mixed populations of free-flying female mosquitoes under semi-field conditions

Great advances in automated identification systems, or 'smart traps', that differentiate insect species have been made in recent years, yet demonstrations of field-ready devices under free-flight conditions remain rare. Here, we describe the results of mixed-species identification of female mosquitoes using an advanced optoacoustic smart trap design under free-flying conditions. Point-of-capture classification was assessed using mixed populations of congeneric (Aedes albopictus and Aedes aegypti) and non-congeneric (Ae. aegypti and Anopheles stephensi) container-inhabiting species of medical importance. Culex quinquefasciatus, also common in container habitats, was included as a third species in all assessments. At the aggregate level, mixed collections of non-congeneric species (Ae. aegypti, Cx. quinquefasciatus, and An. stephensi) could be classified at accuracies exceeding 90% (% error = 3.7-7.1%). Conversely, error rates increased when analysing individual replicates (mean % error = 48.6; 95% CI 8.1-68.6) representative of daily trap captures and at the aggregate level when Ae. albopictus was released in the presence of Ae. aegypti and Cx. quinquefasciatus (% error = 7.8-31.2%). These findings highlight the many challenges yet to be overcome but also the potential operational utility of optoacoustic surveillance in low diversity settings typical of urban environments.

A convolutional neural network to identify mosquito species (Diptera: Culicidae) of the genus Aedes by wing images

Accurate species identification is crucial to assess the medical relevance of a mosquito specimen, but requires intensive experience of the observers and well-equipped laboratories. In this proof-of-concept study, we developed a convolutional neural network (CNN) to identify seven Aedes species by wing images, only. While previous studies used images of the whole mosquito body, the nearly two-dimensional wings may facilitate standardized image capture and reduce the complexity of the CNN implementation. Mosquitoes were sampled from different sites in Germany. Their wings were mounted and photographed with a professional stereomicroscope. The data set consisted of 1155 wing images from seven Aedes species as well as 554 wings from different non-Aedes mosquitoes. A CNN was trained to differentiate between Aedes and non-Aedes mosquitoes and to classify the seven Aedes species based on grayscale and RGB images. Image processing, data augmentation, training, validation and testing were conducted in python using deep-learning framework PyTorch. Our best-performing CNN configuration achieved a macro F1 score of 99% to discriminate Aedes from non-Aedes mosquito species. The mean macro F1 score to predict the Aedes species was 90% for grayscale images and 91% for RGB images. In conclusion, wing images are sufficient to identify mosquito species by CNNs.

An overview of the mosquitoes of Saudi Arabia (Diptera: Culicidae), with updated keys to the adult females

Despite the fact that mosquito-borne infections have considerable consequences for public health in Saudi Arabia, there is neither a thorough review of the species that occur in the country nor updated keys for the identification of the adult females. In this study, species accounts are given for 49 Saudi Arabian mosquito species, as well as Aedes albopictus (Skuse), which is not recorded in Saudi Arabia, but is medically important and is found in some countries of the Middle East and North Africa. Taxonomic notes provide additional information for certain taxa and/or aid their identification.

Automatic identification of medically important mosquitoes using embedded learning approach-based image-retrieval system

Mosquito-borne diseases such as dengue fever and malaria are the top 10 leading causes of death in low-income countries. Control measure for the mosquito population plays an essential role in the fight against the disease. Currently, several intervention strategies; chemical-, biological-, mechanical- and environmental methods remain under development and need further improvement in their effectiveness. Although, a conventional entomological surveillance, required a microscope and taxonomic key for identification by professionals, is a key strategy to evaluate the population growth of these mosquitoes, these techniques are tedious, time-consuming, labor-intensive, and reliant on skillful and well-trained personnel. Here, we proposed an automatic screening, namely the deep metric learning approach and its inference under the image-retrieval process with Euclidean distance-based similarity. We aimed to develop the optimized model to find suitable miners and suggested the robustness of the proposed model by evaluating it with unseen data under a 20-returned image system. During the model development, well-trained ResNet34 are outstanding and no performance difference when comparing five data miners that showed up to 98% in its precision even after testing the model with both image sources: stereomicroscope and mobile phone cameras. The robustness of the proposed-trained model was tested with secondary unseen data which showed different environmental factors such as lighting, image scales, background colors and zoom levels. Nevertheless, our proposed neural network still has great performance with greater than 95% for sensitivity and precision, respectively. Also, the area under the ROC curve given the learning system seems to be practical and empirical with its value greater than 0.960. The results of the study may be used by public health authorities to locate mosquito vectors nearby. If used in the field, our research tool in particular is believed to accurately represent a real-world scenario.

A Swin Transformer-based model for mosquito species identification

Mosquito transmit numbers of parasites and pathogens resulting in fatal diseases. Species identification is a prerequisite for effective mosquito control. Existing morphological and molecular classification methods have evitable disadvantages. Here we introduced Deep learning techniques for mosquito species identification. A balanced, high-definition mosquito dataset with 9900 original images covering 17 species was constructed. After three rounds of screening and adjustment-testing (first round among 3 convolutional neural networks and 3 Transformer models, second round among 3 Swin Transformer variants, and third round between 2 images sizes), we proposed the first Swin Transformer-based mosquito species identification model (Swin MSI) with 99.04% accuracy and 99.16% F1-score. By visualizing the identification process, the morphological keys used in Swin MSI were similar but not the same as those used by humans. Swin MSI realized 100% subspecies-level identification in Culex pipiens Complex and 96.26% accuracy for novel species categorization. It presents a promising approach for mosquito identification and mosquito borne diseases control.

Back to the Future: Quantifying Wing Wear as a Method to Measure Mosquito Age

Vector biologists have long sought the ability to accurately quantify the age of wild mosquito populations, a metric used to measure vector control efficiency. This has proven difficult due to the difficulties of working in the field and the biological complexities of wild mosquitoes. Ideal age grading techniques must overcome both challenges while also providing epidemiologically relevant age measurements. Given these requirements, the Detinova parity technique, which estimates age from the mosquito ovary and tracheole skein morphology, has been most often used for mosquito age grading despite significant limitations, including being based solely on the physiology of ovarian development. Here, we have developed a modernized version of the original mosquito aging method that evaluated wing wear, expanding it to estimate mosquito chronological age from wing scale loss. We conducted laboratory experiments using adult Anopheles gambiae held in insectary cages or mesocosms, the latter of which also featured ivermectin bloodmeal treatments to change the population age structure. Mosquitoes were age graded by parity assessments and both human- and computational-based wing evaluations. Although the Detinova technique was not able to detect differences in age population structure between treated and control mesocosms, significant differences were apparent using the wing scale technique. Analysis of wing images using averaged left- and right-wing pixel intensity scores predicted mosquito age at high accuracy (overall test accuracy: 83.4%, average training accuracy: 89.7%). This suggests that this technique could be an accurate and practical tool for mosquito age grading though further evaluation in wild mosquito populations is required.

Modified Mosquito Programs’ Surveillance Needs and An Image-Based Identification Tool to Address Them

Effective mosquito surveillance and control relies on rapid and accurate identification of mosquito vectors and confounding sympatric species. As adoption of modified mosquito (MM) control techniques has increased, the value of monitoring the success of interventions has gained recognition and has pushed the field away from traditional ‘spray and pray’ approaches. Field evaluation and monitoring of MM control techniques that target specific species require massive volumes of surveillance data involving species-level identifications. However, traditional surveillance methods remain time and labor-intensive, requiring highly trained, experienced personnel. Health districts often lack the resources needed to collect essential data, and conventional entomological species identification involves a significant learning curve to produce consistent high accuracy data. These needs led us to develop MosID: a device that allows for high-accuracy mosquito species identification to enhance capability and capacity of mosquito surveillance programs. The device features high-resolution optics and enables batch image capture and species identification of mosquito specimens using computer vision. While development is ongoing, we share an update on key metrics of the MosID system. The identification algorithm, tested internally across 16 species, achieved 98.4 ± 0.6% % macro F1-score on a dataset of known species, unknown species used in training, and species reserved for testing (species, specimens respectively: 12, 1302; 12, 603; 7, 222). Preliminary user testing showed specimens were processed with MosID at a rate ranging from 181-600 specimens per hour. We also discuss other metrics within technical scope, such as mosquito sex and fluorescence detection, that may further support MM programs.

Application of Deep Learning to Community-Science-Based Mosquito Monitoring and Detection of Novel Species

Mosquito-borne diseases account for human morbidity and mortality worldwide, caused by the parasites (e.g., malaria) or viruses (e.g., dengue, Zika) transmitted through bites of infected female mosquitoes. Globally, billions of people are at risk of infection, imposing significant economic and public health burdens. As such, efficient methods to monitor mosquito populations and prevent the spread of these diseases are at a premium. One proposed technique is to apply acoustic monitoring to the challenge of identifying wingbeats of individual mosquitoes. Although researchers have successfully used wingbeats to survey mosquito populations, implementation of these techniques in areas most affected by mosquito-borne diseases remains challenging. Here, methods utilizing easily accessible equipment and encouraging community-scientist participation are more likely to provide sufficient monitoring. We present a practical, community-science-based method of monitoring mosquito populations using smartphones. We applied deep-learning algorithms (TensorFlow Inception v3) to spectrogram images generated from smartphone recordings associated with six mosquito species to develop a multiclass mosquito identification system, and flag potential invasive vectors not present in our sound reference library. Though TensorFlow did not flag potential invasive species with high accuracy, it was able to identify species present in the reference library at an 85% correct identification rate, an identification rate markedly higher than similar studies employing expensive recording devices. Given that we used smartphone recordings with limited sample sizes, these results are promising. With further optimization, we propose this novel technique as a way to accurately and efficiently monitor mosquito populations in areas where doing so is most critical.

A Computer Vision Approach to Identifying Ticks Related to Lyme Disease

Background: Lyme disease (caused by Borrelia burgdorferi) is an infectious disease transmitted to humans by a bite from infected blacklegged ticks (Ixodes scapularis) in eastern North America. Lyme disease can be prevented if antibiotic prophylaxis is given to a patient within 72 hours of a blacklegged tick bite. Therefore, recognizing a blacklegged tick could facilitate the management of Lyme disease. Methods: In this work, we build an automated detection tool that can differentiate blacklegged ticks from other tick species using advanced computer vision approaches in real-time. Specially, we use convolution neural network models, trained end-to-end, to classify tick species. Also, advanced knowledge transfer techniques are adopted to improve the performance of convolution neural network models. Results: Our best convolution neural network model achieves 92% accuracy on unseen tick species. Conclusion: Our proposed vision-based approach simplifies tick identification and contributes to the emerging work on public health surveillance of ticks and tick-borne diseases. In addition, it can be integrated with the geography of exposure and potentially be leveraged to inform the risk of Lyme disease infection. This is the first report of using deep learning technologies to classify ticks, providing the basis for automation of tick surveillance, and advancing tick-borne disease ecology and risk management.

The Automatic Classification of Pyriproxyfen-Affected Mosquito Ovaries

Pyriproxyfen (PPF) may become an alternative insecticide for areas where pyrethroid-resistant vectors are prevalent. The efficacy of PPF can be assessed through the dissection and assessment of vector ovaries. However, this reliance on expertise is subject to limitations. We show here that these limitations can be overcome using a convolutional neural network (CNN) to automate the classification of egg development and thus fertility status. Using TensorFlow, a resnet-50 CNN was pretrained with the ImageNet dataset. This CNN architecture was then retrained using a novel dataset of 524 dissected ovary images from An. gambiae s.l. An. gambiae Akron, and An. funestus s.l., whose fertility status and PPF exposure were known. Data augmentation increased the training set to 6973 images. A test set of 157 images was used to measure accuracy. This CNN model achieved an accuracy score of 94%, and application took a mean time of 38.5 s. Such a CNN can achieve an acceptable level of precision in a quick, robust format and can be distributed in a practical, accessible, and free manner. Furthermore, this approach is useful for measuring the efficacy and durability of PPF treated bednets, and it is applicable to any PPF-treated tool or similarly acting insecticide.

Identification of public submitted tick images: A neural network approach

Ticks and tick-borne diseases represent a growing public health threat in North America and Europe. The number of ticks, their geographical distribution, and the incidence of tick-borne diseases, like Lyme disease, are all on the rise. Accurate, real-time tick-image identification through a smartphone app or similar platform could help mitigate this threat by informing users of the risks associated with encountered ticks and by providing researchers and public health agencies with additional data on tick activity and geographic range. Here we outline the requirements for such a system, present a model that meets those requirements, and discuss remaining challenges and frontiers in automated tick identification. We compiled a user-generated dataset of more than 12,000 images of the three most common tick species found on humans in the U.S.: Amblyomma americanum, Dermacentor variabilis, and Ixodes scapularis. We used image augmentation to further increase the size of our dataset to more than 90,000 images. Here we report the development and validation of a convolutional neural network which we call "TickIDNet," that scores an 87.8% identification accuracy across all three species, outperforming the accuracy of identifications done by a member of the general public or healthcare professionals. However, the model fails to match the performance of experts with formal entomological training. We find that image quality, particularly the size of the tick in the image (measured in pixels), plays a significant role in the network's ability to correctly identify an image: images where the tick is small are less likely to be correctly identified because of the small object detection problem in deep learning. TickIDNet's performance can be increased by using confidence thresholds to introduce an "unsure" class and building image submission pipelines that encourage better quality photos. Our findings suggest that deep learning represents a promising frontier for tick identification that should be further explored and deployed as part of the toolkit for addressing the public health consequences of tick-borne diseases.

Geometric morphometrics and machine learning as tools for the identification of sibling mosquito species of the Maculipennis complex (Anopheles)

Geometric morphometrics allows researchers to use the specific software to quantify and to visualize morphological differences between taxa from insect wings. Our objective was to assess wing geometry to distinguish four Anopheles sibling species of the Maculipennis complex, An. maculipennis s. s., An. daciae sp. inq., An. atroparvus and An. melanoon, found in Northern Italy. We combined the geometric morphometric approach with different machine learning alghorithms: support vector machine (SVM), random forest (RF), artificial neural network (ANN) and an ensemble model (EN). Centroid size was smaller in An. atroparvus than in An. maculipennis s. s. and An. daciae sp. inq. Principal component analysis (PCA) explained only 33% of the total variance and appeared not very useful to discriminate among species, and in particular between An. maculipennis s. s. and An. daciae sp. inq. The performance of four different machine learning alghorithms using procrustes coordinates of wing shape as predictors was evaluated. All models showed ROC-AUC and PRC-AUC values that were higher than the random classifier but the SVM algorithm maximized the most metrics on the test set. The SVM algorithm with radial basis function allowed the correct classification of 83% of An. maculipennis s. s. and 79% of An. daciae sp. inq. ROC-AUC analysis showed that three landmarks, 11, 16 and 15, were the most important procrustes coordinates in mean wing shape comparison between An. maculipennis s. s. and An. daciae sp. inq. The pattern in the three-dimensional space of the most important procrustes coordinates showed a clearer differentiation between the two species than the PCA. Our study demonstrated that machine learning algorithms could be a useful tool combined with the wing geometric morphometric approach.

Mosquito species identification using convolutional neural networks with a multitiered ensemble model for novel species detection

With over 3500 mosquito species described, accurate species identification of the few implicated in disease transmission is critical to mosquito borne disease mitigation. Yet this task is hindered by limited global taxonomic expertise and specimen damage consistent across common capture methods. Convolutional neural networks (CNNs) are promising with limited sets of species, but image database requirements restrict practical implementation. Using an image database of 2696 specimens from 67 mosquito species, we address the practical open-set problem with a detection algorithm for novel species. Closed-set classification of 16 known species achieved 97.04 ± 0.87% accuracy independently, and 89.07 ± 5.58% when cascaded with novelty detection. Closed-set classification of 39 species produces a macro F1-score of 86.07 ± 1.81%. This demonstrates an accurate, scalable, and practical computer vision solution to identify wild-caught mosquitoes for implementation in biosurveillance and targeted vector control programs, without the need for extensive image database development for each new target region.

Deep learning approaches for challenging species and gender identification of mosquito vectors

Microscopic observation of mosquito species, which is the basis of morphological identification, is a time-consuming and challenging process, particularly owing to the different skills and experience of public health personnel. We present deep learning models based on the well-known you-only-look-once (YOLO) algorithm. This model can be used to simultaneously classify and localize the images to identify the species of the gender of field-caught mosquitoes. The results indicated that the concatenated two YOLO v3 model exhibited the optimal performance in identifying the mosquitoes, as the mosquitoes were relatively small objects compared with the large proportional environment image. The robustness testing of the proposed model yielded a mean average precision and sensitivity of 99% and 92.4%, respectively. The model exhibited high performance in terms of the specificity and accuracy, with an extremely low rate of misclassification. The area under the receiver operating characteristic curve (AUC) was 0.958 ± 0.011, which further demonstrated the model accuracy. Thirteen classes were detected with an accuracy of 100% based on a confusion matrix. Nevertheless, the relatively low detection rates for the two species were likely a result of the limited number of wild-caught biological samples available. The proposed model can help establish the population densities of mosquito vectors in remote areas to predict disease outbreaks in advance.