LocoTrackAI: advanced convolutional neural network-based tool for monitoring locomotor activity in dengue-infected Aedes aegypti mosquitoes

LocoTrackAI: advanced convolutional neural network-based tool for monitoring locomotor activity in dengue-infected Aedes aegypti mosquitoes | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article LocoTrackAI: advanced convolutional neural network-based tool for monitoring locomotor activity in dengue-infected Aedes aegypti mosquitoes Nouman Javed, Adam J. López-Denman , Prasad N. Paradkar , Asim Bhatti This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6067469/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Monitoring the locomotor activity of mosquitoes is vital for understanding their behavioral patterns and role in disease transmission. Studies investigating the locomotor activity of dengue-infected mosquitoes have faced several limitations, including confining mosquitoes within small tubes that restrict natural movement, discontinuous recordings that fail to provide detailed activity patterns, and the lack of open-source tools to effectively monitor mosquito locomotor activity. Here, we present LocoTrackAI, a robust artificial intelligence-based tool that leverages a convolutional neural network (CNN) and a multi-object tracking algorithm to comprehensively analyze the locomotor activity of Aedes aegypti mosquitoes from videos recorded using standard laboratory cages. LocoTrackAI automatically processes video datasets, tracks individual mosquito identities, and provides detailed locomotion results for both individual mosquitoes and group activity, including spatial distributions, movement patterns, heat maps, and activity ratios. The tool features a Skip Frame function to improve computational efficiency, adjustable movement thresholds for customized sensitivity, and a user-friendly interface that supports unsupervised batch processing, ensuring accuracy and flexibility for diverse research applications. The LocoTrackAI achieved 99.91% accuracy with a low centroid detection error of 0.22 pixels across 36,020 frames and demonstrated a 93.23% success rate in reassigning identities during 207 post-occlusion instances. Using LocoTrackAI, we analyzed the locomotor activity of dengue-infected and noninfected mosquitoes across 3.24 million recorded mosquito positions. Results revealed that infected mosquitoes exhibited significantly higher locomotor activity (p = 0.0009), with 95,726 movements (0.30 mean locomotor activity) compared to 42,173 movements (0.13 mean locomotor activity) in noninfected mosquitoes, representing more than 200% of the activity of noninfected mosquitoes. Additionally, spatial analysis indicated a more extensive and uniform distribution for infected mosquitoes, with entropy values of 3.38 for infected and 3.13 for noninfected mosquitoes. These findings suggest that dengue infection increases locomotor activity and spatial exploration, potentially enhancing the mosquitoes' capacity to locate hosts and spread the virus. Future studies could expand on this work by investigating the locomotor effects of other arboviruses and further developing tools to automate the analysis of feeding and other critical behaviors. Infectious Diseases Computational Biology Behavioral Ecology Artificial Intelligence and Machine Learning Entomology Locomotor activity dengue-infected mosquitoes artificial intelligence post-occlusion spatial analysis arboviruses Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Introduction The World Health Organization (WHO) recognizes mosquito-borne diseases, such as malaria, dengue, and Zika virus, as significant public health issues, with a growing number of people being affected globally [ 1 , 2 ]. Despite extensive scientific research and the introduction of various mosquito control measures over the past century, these diseases remain a persistent global health problem [ 3 ]. Studies suggest that the burden of these diseases is expected to increase due to factors that facilitate the transmission of pathogens, including climate change, population growth, urban expansion, poor urban infrastructure, increased international travel, and global trade [ 4 ]. Examining the interactions between mosquitoes and the pathogens they transmit is crucial for improving our understanding of the transmission dynamics and epidemiology of mosquito-borne diseases. Pathogens, including viruses and parasites, are known to influence the behavioral characteristics of their mosquito hosts [ 5 ]. For instance, studies have demonstrated that Zika virus infection increases mosquito locomotion [ 6 ], while decreasing reproductive capacity [ 7 ]. Similarly, West Nile virus infection has been shown to reduce the size of egg rafts [ 8 ] and impair the ability to locate hosts [ 9 ], while Chikungunya virus infection accelerates oviposition time [ 10 ]. Infections caused by mosquito-borne viruses can also affect the mosquito's nervous system, leading to noticeable behavioral changes [ 11 – 13 ]. Parasites, on the other hand, influence host behavior differently depending on their developmental stage. For example, during the sporozoite stage, malaria parasites enhance behaviors such as host-seeking, probing, and feeding, whereas the oocyst stage suppresses these activities. These stage-specific behavioral changes are thought to increase the efficiency of parasite transmission [ 14 , 15 ]. Dengue virus is one of the most significant mosquito-borne viral diseases globally, responsible for an estimated 390 million infections and around 20,000 deaths each year [ 16 ]. In addition to its impact on human health, dengue virus infection has been found to influence various mosquito behaviors. For example, it has been shown to increase host-seeking behavior [ 17 ], prolong feeding duration [ 18 ], extend probing time [ 19 ], enhance the likelihood of refeeding after an interrupted meal (avidity), and reduce reproductive capacity [ 20 ]. Furthermore, dengue-infected mosquitoes may choose oviposition sites farther from their original locations, potentially contributing to greater disease spread [ 21 ]. While much of the research on dengue-related behavioral changes has focused on feeding behaviors, its influence on mosquito locomotion remains insufficiently studied. Previous attempts to investigate locomotor activity in dengue-infected mosquitoes have been limited by methodological challenges. For example, some studies used small tubes measuring just 1 cm x 7 cm to observe movement [ 17 , 22 ], which may have constrained the mosquitoes' natural activity. In another study, locomotor activity was not recorded continuously [ 23 ], capturing one frame every 60 seconds, potentially resulting in gaps in the data. Historically, researchers have relied on manual observation methods to study mosquito behaviors. However, this approach is labor-intensive, prone to errors, and limits the ability to monitor large numbers of mosquitoes simultaneously. Additionally, certain behavioral studies require continuous observation, making the process highly time-intensive and inefficient. In recent years, artificial intelligence (AI) has emerged as a transformative tool for enhancing visualization techniques [ 24 – 27 ]. AI replicates human cognitive processes through various mechanisms integrated within dynamic computing environments [ 28 ]. Previously, artificial intelligence has been applied in mosquito behavioral research, including tasks such as egg counting, larvae counting, and flight analysis [ 29 – 34 ], as well as in neural studies for classifying neural signals [ 35 ]. AI has also demonstrated its utility in evaluating mosquito control strategies [ 36 ]. However, to the best of our knowledge, no open-source AI-based tool is currently available to automatically track the locomotor activity of mosquitoes. Given the above limitations, this study introduces LocoTrackAI, a tool that leverages convolutional neural networks (CNNs) and a multi-object tracking algorithm to automatically track the locomotor activity of Aedes aegypti mosquitoes. LocoTrackAI does not require videos to be captured with sophisticated setups; it performs effectively using videos recorded in standard laboratory cages. The tool processes a folder of videos automatically, without supervision, and generates tracked videos that maintain the identities of individual mosquitoes. Furthermore, it provides detailed locomotor activity results for both individual mosquitoes and groups of mosquitoes, including charts showing changes in distance across frames, locomotion distribution (indicating how many mosquitoes were flying simultaneously across different frames), frames during which mosquitoes were actively moving, the active-to-inactive mosquito ratio, and heat maps showing spatial movement patterns. This study further employs LocoTrackAI to analyze the locomotor activity of dengue-infected Aedes aegypti mosquitoes using continuous video recordings. Plexiglass cages measuring 25 cm x 25 cm x 40 cm were used to facilitate the mosquitoes' natural movement, enabling detailed insights into their behavior through the tracking and analysis capabilities of LocoTrackAI. Materials and methods This study adopted a systematic methodology to monitor locomotor activity in Aedes aegypti mosquitoes. The process began with mosquito rearing under controlled environmental conditions, followed by infecting mosquitoes with dengue virus to study infection-induced behavioral changes. A custom experimental setup was then designed to facilitate precise data acquisition through video recordings. Subsequent steps involved the preparation of training, validation, and test datasets to develop and evaluate the Convolutional Neural Network (CNN) model and LocoTrackAI. The model's training was performed using training data and then validated using validation data. Test data was subsequently used to assess the performance of LocoTrackAI. Detailed locomotor activity analysis was performed to examine key parameters. Finally, the results were interpreted to gain insights into mosquito behavior under the influence of dengue infection, laying the groundwork for future research advancements. These steps and selected parameters, along with a summary of the results, are illustrated in Fig. 1 . Mosquito rearing and dengue infections All experimental procedures were conducted under biosafety level 3 (BSL-3) conditions within the highly secure insectary facilities at the CSIRO Australian Centre of Disease Preparedness (ACDP). These conditions ensured the safe handling and containment of dengue virus serotype 2 (DENV2) to prevent any risk of contamination or environmental exposure. Aedes aegypti mosquitoes used in the study were maintained in a carefully controlled environment with a constant temperature of 27°C and a relative humidity of 70%, conditions that mimic the optimal climate for mosquito survival and activity. A 12-hour light/dark cycle was implemented to simulate natural circadian rhythms, further ensuring that the mosquitoes exhibited typical behavioral patterns. To maintain uniformity and reduce variability, all mosquitoes originated from the same batch of eggs and were reared under similar conditions, ensuring consistency in their behavioral traits. Adult mosquitoes were provided with unrestricted access to a 10% sucrose solution ( ad libitum ), which served as their primary energy source and helped maintain their vitality during the experiments. The DENV2 isolate ET300 (GenBank accession number EF440433) was selected for this study due to its relevance in understanding viral transmission dynamics. Before mosquito exposure, the virus was propagated in Vero cell monolayer cultures to ensure sufficient viral titers for infection. The exposure of Aedes aegypti mosquitoes to DENV followed established protocols [ 37 ]. Female mosquitoes were offered an infectious blood meal using an artificial feeding system, which involved chicken blood and skin to mimic natural feeding behaviors. For the uninfected control groups, female mosquitoes were fed with non-infectious blood to serve as a baseline for comparison in subsequent behavioral analyses. After the feeding process, female mosquitoes were monitored, and those that successfully engorged were carefully selected for experiments. Experimental setup: Infected and uninfected Aedes aegypti mosquitoes were housed in custom-designed rectangular plexiglass cages, each measuring 25 cm x 25 cm x 40 cm. These cages were specifically designed to optimize the observation and recording process, featuring a movable solid white plexiglass wall on one side. This wall was included to minimize background interference, which is often caused by standard netting sleeves, and to provide a clean, unobstructed area for recording. This setup ensured that mosquito movements could be observed and analyzed with precision, particularly in relation to their locomotion behavior analysis. At twelve days post-blood feeding, groups of 5 infected mosquitoes or 5 uninfected mosquitoes were introduced into the cages for separate observation sessions. A single Flea3 camera, manufactured by Point Grey Research in Canada and equipped with infrared capture capabilities, was positioned to record the mosquitoes' movements (Fig. 2 ). This setup allowed for accurate and detailed data collection, facilitating in-depth analysis of locomotion behavior. Data collection: Data collection for the study was performed twelve days post-infection (dpi) by capturing video recordings at a frame rate of 60 frames per second. The process involved three groups, each consisting of five infected mosquitoes, along with three additional groups, each containing five uninfected mosquitoes. The Flea3 camera, which comes with a native resolution of 1280 x 1024 pixels, was reconfigured to a frame size of 1040 x 1024 pixels. This adjustment was made to align the camera's field of view with the dimensions of the mosquito cages and to reduce any unnecessary background from the recordings, ensuring that the focus remained solely on the mosquitoes. Each recording session was conducted independently for each group of infected and noninfected mosquitoes, with a duration of 30 minutes per session, resulting in a total recording time of 180 minutes. Mosquito detection and locomotion tracking: Mosquito detection and locomotion tracking were performed using LocoTrackAI, which is equipped with a convolutional neural network (CNN) and a multi-object tracking algorithm. This tool facilitated the automated analysis of video recordings, ensuring accurate tracking of individual mosquitoes as well as their overall activity within the experimental cages. Details about the CNN model architecture, the multi-object tracking algorithm, and the graphical user interface are provided in the sections below. Model architecture: LocoTrackAI utilizes YOLO11 for object detection (the latest version as of February 2025) [ 38 ], Deep SORT for multi-object tracking [ 39 ], and advanced techniques for analyzing the locomotion of Aedes aegypti mosquitoes. The YOLO11 architecture includes a distinct backbone and head design optimized for real-time detection tasks. The backbone integrates C3K2 blocks, which improve feature extraction by processing smaller feature maps through efficient convolutional layers while preserving essential details. This approach enhances both speed and accuracy compared to earlier YOLO versions. Additionally, the Spatial Pyramid Pooling Fast (SPFF) module aggregates features from different scales, improving the detection of small objects. The C2PSA (Cross Stage Partial with Spatial Attention) block further refines spatial focus, allowing the model to detect critical regions in complex scenes. The detection head employs multi-scale predictions to ensure accurate localization of objects across varying sizes. Deep SORT is employed for multi-object tracking, combining a Kalman filter to predict object states such as position, velocity, and acceleration, with the Hungarian algorithm to match objects efficiently. This process evaluates a cost matrix that incorporates motion consistency through Mahalanobis distance and appearance similarity via cosine distance, enabling robust tracking even during temporary occlusions or overlaps (Fig. 3 a). LocoTrackAI graphical user interface: The tool utilizes a trained model, which can be selected through the "Select Model" button. A pre-trained model is included for immediate application, providing users with a ready-to-use option. Alternatively, users can train their own models by following the straightforward guidelines outlined in [ 38 ]. This feature allows flexibility for both general use and customized applications. Key adjustable parameters available in the tool include Confidence, Intersection over Union (IoU), Frame Size, Threshold, and Skip Frame. The Confidence parameter is critical for filtering detections based on their confidence scores, with only those surpassing a user-specified threshold contributing to the final tracking output. By default, this parameter is set to 0.05, a setting that is particularly suitable for identifying small objects like mosquitoes, ensuring that lower-confidence but potentially relevant detections are included. The IoU parameter is employed during the Non-Maximum Suppression (NMS) stage of the detection process. It assesses the spatial overlap between bounding boxes to remove redundant predictions, thereby enhancing the accuracy of the detection results. The default IoU threshold is also set at 0.05, enabling a more lenient overlap evaluation. This default setting is particularly beneficial for detecting small or closely positioned objects, where a higher threshold might eliminate true-positive detections. The Frame Size parameter allows users to define the maximum input image size, ensuring flexibility across different resolutions. The default value is set to 1040, but LocoTrackAI automatically adjusts this value to the nearest multiple of 32 for compatibility with the neural network architecture. For instance, if the input frame size is 1040, the tool adjusts it to 1056 to optimize processing. The Threshold parameter determines locomotion by calculating the distance between consecutive frame positions of a mosquito in pixels. The tool uses Eq. 1 to classify movement. $$\:\text{Distance}=\sqrt{{\left({x}_{f+1}-{x}_{f}\right)}^{2}+{\left({y}_{f+1}-{y}_{f}\right)}^{2}}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\left(1\right)$$ Where \(\:{x}_{f},\:{y}_{f}\:\) and \(\:{x}_{f+1},\:{y}_{f+1}\:\) represent the coordinates of the mosquito's position in consecutive frames. The resulting distance, measured in pixels, determines whether the movement exceeds the predefined threshold to be classified as locomotion. If the calculated distance exceeds the predefined threshold, the movement is classified as locomotion. The Skip Frame parameter enables users to analyze locomotion at intervals by skipping frames. This is particularly useful when dealing with computational constraints or when studying insects that exhibit slow or infrequent movements. For our analysis, the Skip Frame value was set to 0, meaning no frames were skipped during processing. Setting the value to 1 would skip every alternate frame after processing one frame (e.g., frames 1, 3, 5, etc.), while setting it to 2 would process every third frame (e.g., frames 1, 4, 7, etc.). This feature provides flexibility in balancing computational load and observation detail. After configuring these parameters, users can select a folder containing input videos using the "Select Folder" button. Processing begins automatically upon selection. For each video, LocoTrackAI generates a subfolder in the 'Results' directory (within the input folder) named after the corresponding video. These subfolders contain detailed locomotion analysis results, including Excel files, charts illustrating changes in distance across frames, locomotion distribution (indicating how many insects were moving simultaneously across frames), frames where active movement occurred, the active-to-inactive mosquito ratio, heat maps showing spatial movement patterns, and statistics in text format. LocoTrackAI also allows users to stop processing at any time using the stop button. The tool ensures that any partially processed videos and their results are automatically saved. The graphical user interface of LocoTrackAI is shown in Fig. 4 . Threshold selection for locomotion detection: The threshold for locomotion detection was set at 1.5 pixels, meaning any movement (distance – Eq. 1) greater than this value was classified as locomotion. To determine this threshold, flight video segments of 20 seconds were selected from each sample group, with one segment taken from each of the three infected and three noninfected samples. These segments were processed to identify the average and lowest movement distances when mosquitoes were actively moving. The average movement distance was 4.717 pixels, and the lowest was 1.8 pixels. Based on the lowest recorded value, a threshold of 1.5 pixels was selected to ensure the detection of even minor locomotion. Training, validation and test data: For the customized training of the CNN model and LocoTrackAI performance validation, training, validation, and testing datasets were created using distinct videos to prevent data leakage. Each video, with a duration of 30 minutes, was converted into frames at a rate of 60 frames per second using the Python OpenCV library. For the training dataset, Video 1 from the infected group and Video 1 from the noninfected group were used, from which 4000 frames were randomly selected. The validation dataset was prepared using Video 2 from the infected group and Video 2 from the noninfected group, with 1000 frames randomly selected after frame conversion. As LocoTrackAI takes video sequences as input, the testing dataset was created using Video 3 from the infected group and Video 3 from the noninfected group, where four video sequences were selected, consisting of two sequences from the infected group and two from the noninfected group, with each sequence lasting 30 seconds and the videos having a frame rate of 60 frames per second, resulting in a total of 7204 frames (the extra frame occurs for each video because frame counting starts at 0, not 1). Since each frame contained five mosquitoes, a total of 36,020 mosquito positions were analyzed. This updated dataset split results in 80 percent for training, 20 percent for validation, and a detailed sequence-based testing approach, ensuring robust evaluation and avoiding data leakage. Model training: The CNN model underwent training for 1000 epochs; however, due to the early stopping mechanism with a patience setting of 50, training stopped at epoch 209, as no further improvement in performance was observed. The best performance was recorded at epoch 159. An epoch refers to a complete pass through the training data during the training process. For annotating the training and validation datasets, the Makesense web tool was employed to generate .txt files [ 40 ]. LocoTrackAI evaluation : A. Detection accuracy To evaluate LocoTrackAI, the first step was assessing detection accuracy, which measures how many mosquitoes were correctly detected compared to the actual number present in each frame. Given that there are a total of 7204 test frames across four videos, with five mosquitoes per frame, a total of 36,020 mosquito instances are analyzed for detection accuracy. Detection accuracy was calculated as the ratio of correctly detected mosquitoes to the actual number of mosquitoes in the test frames, where a detection is considered correct if the Intersection over Union (IoU) between the ground truth bounding box and the bounding box calculated by LocoTrackAI is greater than 0.5. This metric provides an overall measure of how well LocoTrackAI detects mosquitoes. The IoU calculation was performed using Python and is presented in Eq. 2 . $$\:IoU=\frac{Area\:of\:intersection}{Area\:of\:union}$$ 2 B. Centroid tracking accuracy: Once detection accuracy was established, the evaluation proceeds to a more detailed analysis by assessing centroid tracking accuracy using the distance error between the ground truth centroid positions and the calculated centroid positions of mosquitoes. Each mosquito was considered an individual instance for evaluation. The Euclidean distance error for each mosquito was computed using Eq. 3. $$\:{E}_{i,f}=\sqrt{{\left({x}_{i,f}-{\widehat{x}}_{i,f}\right)}^{2}+{\left({y}_{i,f}-{\widehat{y}}_{i,f}\right)}^{2}}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\left(3\right)$$ Where \(\:{E}_{i,f}\) represents the distance error for mosquito i in frame f . The coordinates \(\:{x}_{i,f}\) and \(\:{y}_{i,f}\) denote the detected centroid position, while \(\:{\widehat{x}}_{i,f}\) and \(\:{\widehat{y}}_{i,f}\) correspond to the ground truth centroid position of the mosquito in the same frame. This metric helps assess the accuracy of LocoTrackAI in tracking mosquito movements. A detection is considered accurate if the distance error for each mosquito instance is ≤ 1 pixel. If the error exceeds 1 pixel, the detection is classified as inaccurate, indicating potential tracking deviations. This threshold is necessary because mosquitoes are not single-pixel organisms, and their positions vary naturally, making precise centroid estimation inherently challenging. Additionally, manual ground truth annotation may introduce minor inaccuracies, further justifying the need for a reasonable error tolerance in evaluating tracking performance. By analyzing each mosquito individually across 7204 test frames, this approach provided a detailed accuracy evaluation, helping to identify inconsistencies in mosquito movement estimation and ensuring precise tracking performance. For ground truth centroid validation, AstroImageJ was used as it allows pixel-level position inspection with decimal accuracy (Fig. 3 b). This enables precise verification of mosquito positions. C. Post-occlusion identity tracking evaluation: To assess the tool's ability to maintain mosquito identities following occlusions, occlusions were detected using the Intersection over Union (IoU) metric (Eq. 2 ) to evaluate the overlap between bounding boxes identified by the Convolutional Neural Network (CNN) for mosquitoes within video frames. An IoU threshold of 0.25 (25%) was applied, indicating that when the IoU value between the bounding boxes of detected mosquitoes exceeded this threshold, the mosquitoes were classified as being occluded. This threshold, representing partial overlap, was selected to prevent higher thresholds from causing overlapping mosquitoes to be mistakenly perceived as a single entity before being categorized as occluded. To verify identity tracking, occlusions were manually reviewed by analyzing videos. The evaluation was conducted on all six videos (three infected and three noninfected), each lasting 30 minutes. Locomotor activity analysis: The locomotor activity of dengue-infected and noninfected mosquitoes was analyzed using 3 samples for each infected and noninfected, with each sample consisting of 108000 frames. The locomotor activity for each sample was calculated as the number of movements (changes in position) detected between consecutive frames. Average locomotor activity for each frame across all samples in both groups was computed, and this data was used to compare patterns between infected and noninfected mosquitoes. For visualization, smoothed GAM (Generalized Additive Model) values and distribution plots of locomotor activity were generated to provide a comprehensive understanding of differences between the groups. To compare the locomotor activity of infected and noninfected mosquitoes, statistical analysis was performed using the Mann-Whitney test, selected due to the non-normal distribution of the data. For statistical analysis, instead of analyzing locomotor activity on a frame-by-frame basis, batch-wise averages over 1,000 frames were used to reduce noise and variability, ensuring a more stable and representative analysis. As the total number of frames across all samples was 324,000 for infected and 324,000 for noninfected mosquitoes, this batching process resulted in 324 batch averages for each group. These batches were created by calculating the average locomotion values of all mosquitoes within the first 1,000 frames of each sample, followed by the next 1,000 frames, and so on. Shannon entropy was employed to quantify the spatial distribution of mosquitoes in the cage (Fig. 3 c). This metric was calculated to assess how evenly mosquito positions were distributed across spatial sections, providing insight into differences in movement patterns and area utilization between infected and noninfected groups. Occupancy data, representing the number of mosquito positions in each section, were normalized into relative frequencies to account for differences in activity levels. Shannon entropy was then calculated based on these normalized values, with higher entropy indicating a more uniform distribution and lower values reflecting clustering in specific areas. This analysis was performed for both infected and noninfected groups to compare their spatial behavior. Statistical analysis was performed using the Python library SciPy. Computation and programming system: LocoTrackAI was implemented on a computing system featuring an Intel® Core™ i9-13980HX processor with 24 cores, capable of reaching up to 5.6 GHz on select performance cores. The system is equipped with 32GB DDR4-3200 RAM and an NVIDIA RTX 4080 GPU, operating on a Windows 11 Pro 64-bit environment. The tool framework for LocoTrackAI leverages Python 3.8.10 alongside essential scientific computing libraries, including NumPy 1.24.4 and SciPy 1.10.1, for numerical computations. Object detection and tracking are facilitated using Torch 1.9.1 + cu111 and OpenCV 4.10.0.84, ensuring efficient deep learning-based analysis of mosquito locomotion. Results A. CNN model training results for validation data: The object detection CNN model exhibited strong performance in accurately identifying mosquitoes. The validation results demonstrated high evaluation metrics, including a precision of 0.994, a recall of 0.992, and a mean average precision (mAP50) of 0.993. The patterns in precision, recall, and mAP50 values across the training epochs are shown in Fig. 5 a-c, highlighting the model's consistent improvement during training. B. Test data results: LocoTrackAI effectively tracked the majority of mosquito positions across 4 test videos, demonstrating its strong performance and reliability. Out of a total of 36,020 mosquito positions, the tool successfully tracked 35,986 (Table 1 ). The detailed results for Test Video 1 are presented in Fig. 6 , while the data for the remaining test videos are included in the supplementary materials (Supplementary Figs. 1 to 3). For Test Video 1, LocoTrackAI detected 9,001 mosquito positions out of 9,005 across the frames. The average change in distance for an individual mosquito was recorded as 2.77 pixels (average of frames with and without locomotion) (Fig. 6 a). The number of frames in which an individual mosquito exhibited locomotor activity totaled 921 (Fig. 6 b). The cumulative distance change for all five mosquitoes across the frames was measured at 5.17 pixels (Fig. 6 c). Additionally, Fig. 6 d illustrates the group activity in terms of the number of mosquitoes in motion across the frames. Analysis of group activity revealed that a single mosquito was moving in 666 frames, while in 353 frames, two mosquitoes were moving simultaneously (Fig. 6 e). Figure 6 f presents the active-to-inactive mosquito ratio across the frames, providing further insights into the overall group activity patterns. Furthermore, Figs. 6 g and 6 h depict heatmaps that represent the spatial distribution of group activity. Figure 6 g displays the heatmap for an individual mosquito, while Fig. 6 h shows the group heatmap for all mosquitoes, highlighting the number of frames and the percentage of time spent in specific regions of the test arena. All processed test videos are available in the supplementary data (Supplementary Videos 1 to 4). Table 1 LocoTrackAI results for test videos Video Number Total Frames Total Mosquito Positions Mosquito ID Positions Per ID Positions Detected by LocoTrackAI 1 1801 9005 1 1801 1801 2 1801 1801 3 1801 1801 4 1801 1801 5 1801 1797 2 1801 9005 1 1801 1801 2 1801 1797 3 1801 1801 4 1801 1801 5 1801 1801 3 1801 9005 1 1801 1801 2 1801 1801 3 1801 1801 4 1801 1779 5 1801 1801 4 1801 9005 1 1801 1801 2 1801 1801 3 1801 1801 4 1801 1797 5 1801 1801 Total 7204 36020 20 36020 35986 Performance validation LocoTrackAI's performance was validated by assessing its effectiveness on 4 test videos using detection accuracy, centroid tracking accuracy, and performing post-occlusion identity tracking evaluation, which are detailed below. A. Detection accuracy: The performance of LocoTrackAI on 4 test videos demonstrated exceptional accuracy in tracking mosquito positions. For most individual mosquitoes, the tool achieved an accuracy of 100%, with the lowest accuracy observed being 98.78% for Mosquito 4 in Video 3. The combined accuracy for all mosquitoes across the test videos was above 99%, with the lowest recorded combined accuracy being 99.76% for Video 3. The average accuracy across all test videos was 99.91% (Table 2 ). These results emphasize the reliability and robustness of LocoTrackAI in tracking mosquitoes. Table 2 LocoTrackAI evaluation results for detecting mosquito positions Video Number Total Frames Mosquito ID Positions Per ID Correctly Detected Positions by LocoTrackAI Incorrect/ Missing Detections Accuracy Per ID Combined Accuracy 1 1801 1 1801 1801 0 100% 99.96% 2 1801 1801 0 100% 3 1801 1801 0 100% 4 1801 1801 0 100% 5 1801 1797 4 99.78% 2 1801 1 1801 1801 0 100% 99.96% 2 1801 1797 4 99.78% 3 1801 1801 0 100% 4 1801 1801 0 100% 5 1801 1801 0 100% 3 1801 1 1801 1801 0 100% 99.76% 2 1801 1801 0 100% 3 1801 1801 0 100% 4 1801 1779 22 98.78% 5 1801 1801 0 100% 4 1801 1 1801 1801 0 100% 99.96% 2 1801 1801 0 100% 3 1801 1801 0 100% 4 1801 1797 4 99.78% 5 1801 1801 0 100% Total/Average 7204 36020 35986 34 99.91% B. Centroid tracking accuracy: Centroid Tracking Accuracy was calculated using the Euclidean distance error, which accounted for errors across both the x and y axes. Figure 7 highlights the mean Euclidean distance error for each frame; for instance, for frame 1, it represents the mean Euclidean distance error of all 20 mosquito positions in that specific frame. To enhance the clarity of the data, a 50-frame moving average was applied, thereby facilitating better understanding of the chart. The overall average of the mean Euclidean distance errors across all test video frames was 0.22 pixels, highlighting the tool's exceptional performance. C. Post-occlusion identity tracking evaluation The LocoTrackAI demonstrated outstanding performance in post-occlusion identity tracking, accurately reassigning identities in 193 out of 207 occlusion instances and achieving a success rate of 93.23%. These results highlight the tool's robustness in maintaining accurate identity tracking while tracking locomotor activity. Locomotor activity analysis The locomotor activity of dengue-infected and noninfected mosquitoes was analyzed across 3 samples of each group, with each sample consisting of 108,000 frames and a total of 540,000 mosquito positions. The cumulative mosquito positions were 1.62 million for infected and 1.62 million for noninfected mosquitoes, amounting to 3.24 million positions in total. The detection accuracy for all samples exceeded 99%, ensuring the reliability of the results (Table 3 ). The analysis revealed that dengue-infected mosquitoes exhibited significantly higher locomotor activity compared to their noninfected counterparts. Among the infected samples, the highest locomotor activity was observed in sample 3, with 55,998 movements, while the lowest was recorded in sample 2, with 7,664 movements. The total locomotor activity across all 3 infected samples was 95,726 movements. In contrast, the noninfected samples showed their highest locomotor activity in sample 2, with 19,187 movements, and their lowest in sample 1, with 5,467, resulting in a total of 42,173 movements across all 3 samples. Table 3 Locomotor activity results for infected and noninfected mosquitoes Sample Number Total Frames Total Positions Positions Detected by LocoTrackAI Detection Accuracy Locomotion Infected 1 108000 540000 539089 99.83% 55998 2 108000 540000 539871 99.98% 7664 3 108000 540000 539611 99.93% 32064 Total/Average 324000 1620000 1618571 99.91% 95726 Noninfected 1 108000 540000 539634 99.93% 5467 2 108000 540000 539702 99.94% 19187 3 108000 540000 539005 99.82% 17519 Total/Average 324000 1620000 1618341 99.90% 42173 Figure 8 presents the results for the first infected and noninfected mosquito samples, while the results for the remaining samples are provided as supplementary material (Supplementary Figs. 4 to 7). Figure 8 a illustrates the change in distance for infected mosquitoes in sample 1, which is 3.67 pixels, and Fig. 8 b depicts the number of frames with locomotion for the same sample. Figure 8 c shows the number of mosquitoes flying simultaneously across different frames for infected mosquitoes in sample 1, where no mosquitoes were moving for 59829 frames, one mosquito was moving for 41155 frames, two mosquitoes were moving for 6281 frames, three mosquitoes were moving for 659 frames, and four mosquitoes were moving for 76 frames. Figure 8 d provides the heatmap for infected mosquitoes in sample 1. For the noninfected mosquitoes, Fig. 8 e displays the change in distance, which is 1.37 pixels, while Fig. 8 f represents the frames with locomotion for noninfected mosquitoes in sample 1. Figure 8 g shows the number of mosquitoes flying simultaneously across different frames for noninfected mosquitoes in sample 1, where no mosquitoes were moving for 102771 frames, one mosquito was moving for 4991 frames, and tow mosquitoes were moving for 238 frames. Finally, Fig. 8 h presents the heatmap for noninfected mosquitoes in sample 1. Figure 9 a illustrates the average locomotor activity of infected and noninfected mosquitoes across all frames, calculated by averaging the locomotor activity for each frame across all samples in each group. To provide a clearer representation, Fig. 9 b displays the smoothed GAM values for both groups. Figure 9 c presents the distribution of average locomotor activity across all samples, demonstrating that infected mosquitoes exhibited higher locomotor activity compared to noninfected mosquitoes. The mean locomotor activity of infected mosquitoes was 0.30 movements, while for noninfected mosquitoes, it was 0.13 movements. Figure 9 d further explores the locomotor activity by presenting values in the form of averages calculated over batches of 1,000 frames each. This data was used to perform a Mann-Whitney test, which yielded a p-value of 0.0009, confirming a statistically significant difference in locomotor activity between the infected and noninfected mosquitoes. These findings highlight the significant impact of dengue infection on the movement patterns of mosquitoes. The spatial distribution patterns of mosquitoes offer important insights into their behavioral tendencies and environmental interaction. Scatterplots presented in Figs. 9 e and 9 f show the spatial positions of infected and noninfected mosquitoes, respectively, illustrating their movement paths within the cage. To further clarify these patterns, Figs. 9 g and 9 h include heatmaps that represent the density of mosquito positions. The Shannon entropy values provide a systematic approach to analyzing these spatial patterns. Infected mosquitoes recorded a higher entropy value of 3.38, indicating a more uniform and extensive spatial distribution across the cage. This suggests that infected mosquitoes explore a broader range of locations within the cage. In contrast, the noninfected mosquitoes demonstrated a lower entropy value of 3.13, reflecting a more concentrated spatial distribution with activity focused in specific areas. This implies that noninfected mosquitoes engage in less exploration and exhibit a preference for localized regions. These results highlight the substantial effect of dengue infection on mosquito behavior. Infected mosquitoes exhibit increased spatial exploration and movement, potentially linked to behaviors such as enhanced host-seeking. Meanwhile, noninfected mosquitoes display more constrained spatial activity and concentrated movement, illustrating a clear behavioral distinction between the two groups. Discussion The LocoTrackAI tool can facilitate a detailed analysis of the locomotor activity of Aedes aegypti mosquitoes. Data generated from LocoTrackAI can be used to explore various aspects of locomotor behavior, such as activity patterns during different times of the day, movement distances, and group dynamics. Furthermore, given the tool's exceptional ability to analyze the complex movement behaviors of small organisms like mosquitoes, it presents significant potential for broader applications, including studies of locomotor activity in other species, such as flies, ticks, and zebrafish. LocoTrackAI demonstrated a high level of detection accuracy, achieving a rate of 99.91%, along with a low centroid detection Euclidean distance error of 0.22 pixels, both validated on 36,020 frames. These results highlight the tool's precision and reliability in tracking mosquito locomotor activity. In addition to its strong performance, the tool provides a user-friendly interface that simplifies video analysis and supports batch processing without the need for constant supervision. Furthermore, LocoTrackAI includes a Skip Frame function to improve computational efficiency by analyzing frames at predefined intervals. It also incorporates a threshold-based classification method, allowing users to adjust sensitivity based on the organism's activity level or specific study requirements. These features collectively enhance the tool's precision, adaptability, and utility for a wide range of research applications. LocoTrackAI has demonstrated strong performance in tracking the locomotor activity of mosquitoes from videos recorded at 60 frames per second, even in the presence of occlusions. However, its accuracy in maintaining correct identities may decrease when using videos with lower frame rates. Previous studies on animal tracking have recommended frame rates above 25 frames per second for organisms such as mice, zebrafish, and ants [ 41 , 42 ]. Given the small size, fast movements, and complex locomotor patterns of mosquitoes, higher frame rates may be required compared to those used for tracking the locomotor activity of mice, zebrafish, and ants. Additionally, gaps in data points caused by background distortions or light reflections could reduce tracking performance. To address this, using a clean and uniform background, as demonstrated with the simple white wall in this study, is recommended. Increasing the size of the training dataset and employing anti-glare materials such as plexiglass can further improve performance by reducing environmental distractions and enhancing the quality of input data for locomotor activity analysis. This study highlights the significant impact of dengue infection on the locomotor activity of Aedes aegypti mosquitoes. Dengue-infected mosquitoes demonstrated substantially higher activity levels compared to their noninfected counterparts, with more frequent movements and greater spatial exploration. These behavioral changes may enable them to cover larger areas, encounter more hosts [ 43 ], and locate additional breeding sites [ 21 ], which may increase their potential for virus transmission. In contrast, noninfected mosquitoes displayed lower activity levels and a more localized spatial distribution, suggesting restricted movement and energy conservation as a baseline behavioral pattern. This study also confirms previously observed increases in the locomotor activity of dengue-infected mosquitoes while addressing limitations of earlier studies, such as the use of small confinement tubes (1 cm x 7 cm) [ 17 , 22 ] and discontinuous recordings [ 23 ]. By utilizing continuous video recordings in larger plexiglass cages (25 cm x 25 cm x 40 cm), this study provides a more natural and comprehensive analysis of movement patterns. Overall, these findings emphasize how dengue infection modulates mosquito behavior in ways that may enhance their vectorial capacity, offering valuable perspectives for understanding disease dynamics and informing vector control strategies. Conclusions In conclusion, this study demonstrates the effectiveness of LocoTrackAI in analyzing the locomotor activity of Aedes aegypti mosquitoes and highlights the significant behavioral changes induced by dengue infection. LocoTrackAI achieved a high accuracy rate of 99.91% with a low centroid detection error of 0.22 pixels, validated on 36,020 frames. Its advanced features, such as batch processing, Skip Frame functionality, and adjustable thresholds, make LocoTrackAI a reliable tool for analyzing mosquito locomotor activity and a valuable resource for studying mosquito-pathogen interactions and vector control strategies. The behavioral analysis revealed that dengue-infected mosquitoes exhibited significantly higher locomotor activity, with a total of 95,726 movements across samples, compared to 42,173 movements in noninfected mosquitoes. The mean locomotor activity of infected mosquitoes was 0.30 movements, representing over 200% of the 0.15 observed in noninfected mosquitoes. Infected mosquitoes also showed broader spatial exploration, with higher entropy values (3.38 compared to 3.13 in noninfected mosquitoes), indicating a more uniform distribution across the cage. These behavioral differences suggest that infected mosquitoes have an increased capacity to locate hosts and spread the virus, enhancing their transmission potential. Future research could expand on this work in several ways. On the behavioral side, studies could explore the locomotor activity of mosquitoes infected with other arboviruses to determine whether the observed changes are specific to dengue or represent a broader trend. On the technological side, tools like LocoTrackAI could be further developed to automate the analysis of other critical behaviors, such as feeding activity and feeding duration, providing a more comprehensive understanding of mosquito-pathogen interactions. These advancements would contribute significantly to improving vector-borne disease models and developing more targeted mosquito control strategies. Declarations Acknowledgement: The authors acknowledge the capabilities of the CSIRO Australian Centre for Disease Preparedness ( grid.413322.5 ) in undertaking this research, including infrastructure by the National Collaborative Research Infrastructure Strategy (NCRIS). Animal ethics statement: The experiments for chicken blood collection were conducted under animal ethics guidelines with approval from ACDP Animal Ethics Committee (#ACDP22010). Authors' contributions: Writing—original draft preparation, N.J.; writing—review and editing, A.B., A.J.L., and P.N.P.; visualization, N.J.; Software, N.J.; Data curation, N.J., A.J.L.; supervision, A.B., A.J.L., and P.N.P. All authors have read and agreed to the published version of the manuscript. Competing interests: The authors declare no competing interests. Funding declaration: The study was partly funded by CSIRO, which provided strategic funding to P.N.P. and N.J. The funders had no role in study design, data collection and analysis, decision to publish, or manuscript preparation. Data availability: Supplementary images, supplementary videos, videos for testing, and LocoTrackAI graphical user interface .exe file is available at: https://drive.google.com/drive/folders/1bxgp0GasgOswm0Ox7E71Yt6yCPtBKRUM?usp=sharing References Organization, W.H., Global technical strategy for malaria 2016-2030 . 2015: World Health Organization. World Health Organization. Vector-borne diseases . 2024; Available from: https://www.who.int/news-room/fact-sheets/detail/vector-borne-diseases. World Health Organization. Malaria in children under five . 2018; Available from: https://www.who.int/malaria/areas/high_risk_groups/children/en/. Franklinos, L.H., et al., The effect of global change on mosquito-borne disease. The Lancet Infectious Diseases, 2019. 19 (9): p. e302-e312. Javed, N., A. Bhatti, and P.N. Paradkar, Advances in Understanding Vector Behavioural Traits after Infection. Pathogens, 2021. 10 (11): p. 1376. Padilha, K.P., et al., Zika infection decreases Aedes aegypti locomotor activity but does not influence egg production or viability. Memórias do Instituto Oswaldo Cruz, 2018. 113 . Petersen, M.T., et al., The impact of the age of first blood meal and Zika virus infection on Aedes aegypti egg production and longevity. PLoS One, 2018. 13 (7): p. e0200766. Styer, L.M., M.A. Meola, and L.D. Kramer, West Nile Virus Infection Decreases Fecundity of Culex tarsalis Females. Journal of Medical Entomology, 2007. 44 (6): p. 1074-1085. Vogels, C.B.F., et al., Virus interferes with host-seeking behaviour of mosquito. The Journal of experimental biology, 2017. 220 (Pt 19): p. 3598-3603. Martin, E., et al., Differential responses of the mosquito Aedes albopictus from the Indian Ocean region to two chikungunya isolates. BMC Ecol, 2010. 10 : p. 8. Bhatti, A., et al., Emerging Trends in Neuro Engineering and Neural Computation . 2017: Springer. Gaburro, J., et al., Zika virus-induced hyper excitation precedes death of mouse primary neuron. Virology journal, 2018. 15 (1): p. 1-13. Gaburro, J., S. Nahavandi, and A. Bhatti, Insects Neural Model: Potential Alternate to Mammals for Electrophysiological Studies , in Emerging Trends in Neuro Engineering and Neural Computation . 2017, Springer. p. 119-130. Cator, L.J., et al., 'Manipulation' without the parasite: altered feeding behaviour of mosquitoes is not dependent on infection with malaria parasites. Proceedings of the Royal Society B: Biological Sciences, 2013. 280 (1763): p. 20130711. Koella, J.C., L. Rieu, and R.E.L. Paul, Stage-specific manipulation of a mosquito's host-seeking behavior by the malaria parasite Plasmodium gallinaceum. Behavioral Ecology, 2002. 13 (6): p. 816-820. Brady, O.J., et al., Refining the global spatial limits of dengue virus transmission by evidence-based consensus. 2012. Tallon, A.K., et al., Dengue infection modulates locomotion and host seeking in Aedes aegypti. PLoS neglected tropical diseases, 2020. 14 (9): p. e0008531-e0008531. Sim, S., J.L. Ramirez, and G. Dimopoulos, Dengue virus infection of the Aedes aegypti salivary gland and chemosensory apparatus induces genes that modulate infection and blood-feeding behavior. PLoS Pathog, 2012. 8 (3): p. e1002631. Sylvestre, G., M. Gandini, and R. Maciel-de-Freitas, Age-dependent effects of oral infection with dengue virus on Aedes aegypti (Diptera: Culicidae) feeding behavior, survival, oviposition success and fecundity. PLoS One, 2013. 8 (3): p. e59933. Maciel-de-Freitas, R., et al., The Influence of Dengue Virus Serotype-2 Infection on Aedes aegypti (Diptera: Culicidae) Motivation and Avidity to Blood Feed. PLOS ONE, 2013. 8 (6): p. e65252. Gaburro, J., et al., Dengue virus infection changes Aedes aegypti oviposition olfactory preferences. Scientific Reports, 2018. 8 (1): p. 13179. Lima-Camara, T.N., et al., Dengue infection increases the locomotor activity of Aedes aegypti females. PloS one, 2011. 6 (3): p. e17690-e17690. Gaburro, J., et al., Neurotropism and behavioral changes associated with Zika infection in the vector Aedes aegypti. Emerg Microbes Infect, 2018. 7 (1): p. 68. AlShamsi, M., et al., Artificial intelligence and blockchain for transparency in governance , in Artificial Intelligence for Sustainable Development: Theory, Practice and Future Applications . 2021, Springer. p. 219-230. Matthews, G., et al., Evolution and revolution: Personality research for the coming world of robots, artificial intelligence, and autonomous systems. Personality and individual differences, 2021. 169 : p. 109969. Nawaz, M.S., et al., Using artificial intelligence techniques for COVID-19 genome analysis. Applied Intelligence, 2021: p. 1-18. Yang, S.J., et al., Human-centered artificial intelligence in education: Seeing the invisible through the visible. Computers and Education: Artificial Intelligence, 2021. 2 : p. 100008. Shabbir, J. and T. Anwer, Artificial intelligence and its role in near future. arXiv preprint arXiv:1804.01396, 2018. Javed, N., et al., EggCountAI: a convolutional neural network-based software for counting of Aedes aegypti mosquito eggs. Parasites & Vectors, 2023. 16 (1): p. 341. Javed, N., et al., FlightTrackAI: a robust convolutional neural network-based tool for tracking the flight behaviour of Aedes aegypti mosquitoes. Royal Society Open Science, 2024. 11 (10): p. 240923. Javed, N., et al., LarvaeCountAI: a robust convolutional neural network-based tool for accurately counting the larvae of Culex annulirostris mosquitoes. Acta Tropica, 2024. 260 : p. 107468. Javed, N., et al., Flight traits of dengue-infected Aedes aegypti mosquitoes. Computers in Biology and Medicine, 2024. 171 : p. 108178. Javed, N., P.N. Paradkar, and A. Bhatti, Flight behaviour monitoring and quantification of aedes aegypti using convolution neural network. PloS one, 2023. 18 (7): p. e0284819. Javed, N., P.N. Paradkar, and A. Bhatti, An overview of technologies available to monitor behaviours of mosquitoes. Acta Tropica, 2024. 258 : p. 107347. Sharifrazi, D., et al., Automated detection of Zika and dengue in Aedes aegypti using neural spiking analysis: A machine learning approach. Biomedical Signal Processing and Control, 2024. 96 : p. 106594. Joshi, A. and C. Miller, Review of machine learning techniques for mosquito control in urban environments. Ecological Informatics, 2021. 61 : p. 101241. Duchemin, J.-B., et al., Zika vector transmission risk in temperate Australia: a vector competence study. Virology journal, 2017. 14 (1): p. 1-10. Jocher, G., A. Chaurasia, and J. Qiu. Ultralytics YOLO . 2023; Available from: https://github.com/ultralytics/ultralytics. Wojke, N., A. Bewley, and D. Paulus. Simple online and realtime tracking with a deep association metric . in 2017 IEEE international conference on image processing (ICIP) . 2017. IEEE. Piotr Skalski. Make Sense . 2019; Available from: https://www.makesense.ai. Pérez-Escudero, A., et al., idTracker: tracking individuals in a group by automatic identification of unmarked animals. Nature Methods, 2014. 11 (7): p. 743-748. Romero-Ferrero, F., et al., idtracker.ai: tracking all individuals in small or large collectives of unmarked animals. Nature Methods, 2019. 16 (2): p. 179-182. Evans, O., et al., Increased locomotor activity and metabolism of Aedes aegypti infected with a life-shortening strain of Wolbachia pipientis. Journal of Experimental Biology, 2009. 212 (10): p. 1436-1441. Additional Declarations The authors declare no competing interests. Supplementary Files SupplementaryData.zip Supplementary Data Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6067469","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":418272553,"identity":"85a1aaa9-6114-441a-b198-2d584c11d0cd","order_by":0,"name":"Nouman Javed","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7UlEQVRIiWNgGAWjYHCCBCBmlmM4wMAMZgIZxGkxJkkLCDAnNoC0MBCjxZyB4eFnngrr9L7jzY8NHu5gkOO7kcD4mQePFssGhmRpnjPpuTPPHDNOSDzDYCx5I4FZGp8WgwMMCZIz2w7nbriRYHwgsY0hEchgIKQl+efMf4fTDe4//wzSUg/UwvybgJY0iY8NhxMMbvAAHdbGAGQksOG35TBDmsWHY+mGM8/kFBsktkkAGQ/bLOfg03K8J/lGQo21PN/x45slf7bZABnJh2+8waOFgZknAZkrAcSMDfg0AAH7AQIKRsEoGAWjYMQDAIZ6Uj/qc/cLAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-0520-3504","institution":"Institute for Intelligent Systems Research and Innovation, Deakin University, Geelong, Victoria, 3216 Australia","correspondingAuthor":true,"prefix":"","firstName":"Nouman","middleName":"","lastName":"Javed","suffix":""},{"id":418273325,"identity":"f2572291-28ac-4245-b12c-b88959fb04e9","order_by":1,"name":"Adam J. López-Denman ","email":"","orcid":"https://orcid.org/0000-0001-6051-6381","institution":"CSIRO Health \u0026 Biosecurity, Australian Centre for Disease Preparedness, Geelong, Victoria 3220 Australia","correspondingAuthor":false,"prefix":"","firstName":"Adam","middleName":"J.","lastName":"López-Denman ","suffix":""},{"id":418273326,"identity":"475c0540-5e74-48e7-a873-3b6c493e802c","order_by":2,"name":"Prasad N. Paradkar ","email":"","orcid":"https://orcid.org/0000-0002-6553-2214","institution":"CSIRO Health \u0026 Biosecurity, Australian Centre for Disease Preparedness, Geelong, Victoria 3220 Australia","correspondingAuthor":false,"prefix":"","firstName":"Prasad","middleName":"N.","lastName":"Paradkar ","suffix":""},{"id":418273327,"identity":"b8959562-c4ff-4b8f-a955-eb9e77ca4bfa","order_by":3,"name":"Asim Bhatti","email":"","orcid":"https://orcid.org/0000-0001-6876-1437","institution":"Institute for Intelligent Systems Research and Innovation, Deakin University, Geelong, Victoria, 3216 Australia","correspondingAuthor":false,"prefix":"","firstName":"Asim","middleName":"","lastName":"Bhatti","suffix":""}],"badges":[],"createdAt":"2025-02-20 01:06:54","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-6067469/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6067469/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":77022953,"identity":"bee812a0-5c8f-402c-8583-7e915179c04f","added_by":"auto","created_at":"2025-02-24 11:08:23","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":664118,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eWorkflow for the study\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/bc37dfa5332f6deed468cfde.png"},{"id":77021852,"identity":"ee22dba9-8b16-4f8b-9ad8-4bb51c56acea","added_by":"auto","created_at":"2025-02-24 11:00:23","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":877212,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExperimental setup\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/b1d329c25ca0e9cd7c432b90.png"},{"id":77022956,"identity":"54813cdd-fe11-4c86-965c-82d9dad3934b","added_by":"auto","created_at":"2025-02-24 11:08:23","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":567722,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMethodology.\u003c/strong\u003e (a) LocoTrackAI's process for tracking mosquitoes and producing results. (b) Calculation of ground truth centroids using the AstroImageJ tool. (c) Shannon entropy calculation procedure.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/727e5097d3bd96fd6d169acf.png"},{"id":77021855,"identity":"fd693a1c-2a0f-420e-8a3f-a8d6af7732c9","added_by":"auto","created_at":"2025-02-24 11:00:23","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":177375,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGraphical user interface of LocoTrackAI\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/d41acc7608fb7e1eed9c5175.png"},{"id":77021857,"identity":"08d1c526-a253-49a8-b324-fd5a050f8b3e","added_by":"auto","created_at":"2025-02-24 11:00:23","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":146624,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCNN training results on validation data. \u003c/strong\u003e(a) Recall. (b) Precision. (c) mAP50.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/88e9f84f4d91825854e0524c.png"},{"id":77021861,"identity":"91711593-8fb2-4617-b388-4045378e2e88","added_by":"auto","created_at":"2025-02-24 11:00:23","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":857827,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLocoTrackAI test data results. \u003c/strong\u003e(a) Change in distance for an individual mosquito. (b) Frames with individual mosquito locomotion. (c) Change in distance for a group of mosquitoes. (d) Frames with group locomotion. (e) Group activity statistics. (f) Active-to-inactive mosquito ratio across the frames. (g) Heatmap for an individual mosquito. (h) Heatmap for a group of mosquitoes.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/c8c27a752d2b2b8aca0dee14.png"},{"id":77021860,"identity":"6611036f-6842-45b7-9b20-b1f40ef5c41c","added_by":"auto","created_at":"2025-02-24 11:00:23","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":40288,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMean Euclidean distance error across the frames.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/d94570251a33fd065d86efb5.png"},{"id":77022968,"identity":"03033da5-15bf-4a42-8895-e4d58188dda9","added_by":"auto","created_at":"2025-02-24 11:08:24","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":627651,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eInfected and noninfected mosquitoes' locomotion results. \u003c/strong\u003e(a) Change in distance for infected mosquitoes' sample 1. (b) Frames with locomotion for infected mosquitoes' sample 1. (c) Locomotor activity statistics for infected mosquitoes' sample 1. (d) Heatmap for infected mosquitoes' sample 1. (e) Change in distance for noninfected mosquitoes' sample 1. (f) Frames with locomotion for noninfected mosquitoes' sample 1. (g) Locomotor activity statistics for noninfected mosquitoes' sample 1. (h) Heatmap for noninfected mosquitoes' sample 1.\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/14859f7f729bb6f78a1fee5f.png"},{"id":77024858,"identity":"ec1ddeee-99cd-453b-b77b-d11f2938664e","added_by":"auto","created_at":"2025-02-24 11:16:23","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":1049584,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLocomotor activity and spatial distribution of infected and noninfected mosquitoes. \u003c/strong\u003e(a) Average locomotor activity across all frames for infected and noninfected mosquitoes. (b) Smoothed GAM values for infected and noninfected mosquitoes. (c) Distribution of average locomotor activity across all infected and noninfected samples. (d) Distribution of locomotor activity averages calculated over batches of 1,000 frames for infected and noninfected samples. (e) Scatterplot of spatial positions for infected mosquitoes. (f) Scatterplot of spatial positions for noninfected mosquitoes. (g) Heatmap of position density for infected mosquitoes. (h) Heatmap of position density for noninfected mosquitoes.\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/195881b529bc6b63ed90585d.png"},{"id":77026113,"identity":"0668a4e1-ce38-4880-bcbd-b0ea12d3f8f1","added_by":"auto","created_at":"2025-02-24 11:24:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6031598,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/edf0e3de-787e-448b-b7f2-1051fa849db9.pdf"},{"id":77021883,"identity":"6c1b3191-86c2-454d-949f-f835c05fc481","added_by":"auto","created_at":"2025-02-24 11:00:24","extension":"zip","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":63189633,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Data\u003c/p\u003e","description":"","filename":"SupplementaryData.zip","url":"https://assets-eu.researchsquare.com/files/rs-6067469/v1/369ea579aa591810c41d6b3f.zip"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eLocoTrackAI: advanced convolutional neural network-based tool for monitoring locomotor activity in dengue-infected \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eAedes aegypti\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e mosquitoes\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe World Health Organization (WHO) recognizes mosquito-borne diseases, such as malaria, dengue, and Zika virus, as significant public health issues, with a growing number of people being affected globally [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Despite extensive scientific research and the introduction of various mosquito control measures over the past century, these diseases remain a persistent global health problem [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Studies suggest that the burden of these diseases is expected to increase due to factors that facilitate the transmission of pathogens, including climate change, population growth, urban expansion, poor urban infrastructure, increased international travel, and global trade [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eExamining the interactions between mosquitoes and the pathogens they transmit is crucial for improving our understanding of the transmission dynamics and epidemiology of mosquito-borne diseases. Pathogens, including viruses and parasites, are known to influence the behavioral characteristics of their mosquito hosts [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. For instance, studies have demonstrated that Zika virus infection increases mosquito locomotion [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], while decreasing reproductive capacity [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Similarly, West Nile virus infection has been shown to reduce the size of egg rafts [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] and impair the ability to locate hosts [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e], while Chikungunya virus infection accelerates oviposition time [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Infections caused by mosquito-borne viruses can also affect the mosquito's nervous system, leading to noticeable behavioral changes [\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Parasites, on the other hand, influence host behavior differently depending on their developmental stage. For example, during the sporozoite stage, malaria parasites enhance behaviors such as host-seeking, probing, and feeding, whereas the oocyst stage suppresses these activities. These stage-specific behavioral changes are thought to increase the efficiency of parasite transmission [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDengue virus is one of the most significant mosquito-borne viral diseases globally, responsible for an estimated 390\u0026nbsp;million infections and around 20,000 deaths each year [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. In addition to its impact on human health, dengue virus infection has been found to influence various mosquito behaviors. For example, it has been shown to increase host-seeking behavior [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], prolong feeding duration [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], extend probing time [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e], enhance the likelihood of refeeding after an interrupted meal (avidity), and reduce reproductive capacity [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Furthermore, dengue-infected mosquitoes may choose oviposition sites farther from their original locations, potentially contributing to greater disease spread [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. While much of the research on dengue-related behavioral changes has focused on feeding behaviors, its influence on mosquito locomotion remains insufficiently studied. Previous attempts to investigate locomotor activity in dengue-infected mosquitoes have been limited by methodological challenges. For example, some studies used small tubes measuring just 1 cm x 7 cm to observe movement [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e], which may have constrained the mosquitoes' natural activity. In another study, locomotor activity was not recorded continuously [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], capturing one frame every 60 seconds, potentially resulting in gaps in the data.\u003c/p\u003e \u003cp\u003eHistorically, researchers have relied on manual observation methods to study mosquito behaviors. However, this approach is labor-intensive, prone to errors, and limits the ability to monitor large numbers of mosquitoes simultaneously. Additionally, certain behavioral studies require continuous observation, making the process highly time-intensive and inefficient. In recent years, artificial intelligence (AI) has emerged as a transformative tool for enhancing visualization techniques [\u003cspan additionalcitationids=\"CR25 CR26\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. AI replicates human cognitive processes through various mechanisms integrated within dynamic computing environments [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Previously, artificial intelligence has been applied in mosquito behavioral research, including tasks such as egg counting, larvae counting, and flight analysis [\u003cspan additionalcitationids=\"CR30 CR31 CR32 CR33\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], as well as in neural studies for classifying neural signals [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. AI has also demonstrated its utility in evaluating mosquito control strategies [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. However, to the best of our knowledge, no open-source AI-based tool is currently available to automatically track the locomotor activity of mosquitoes.\u003c/p\u003e \u003cp\u003eGiven the above limitations, this study introduces LocoTrackAI, a tool that leverages convolutional neural networks (CNNs) and a multi-object tracking algorithm to automatically track the locomotor activity of \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes. LocoTrackAI does not require videos to be captured with sophisticated setups; it performs effectively using videos recorded in standard laboratory cages. The tool processes a folder of videos automatically, without supervision, and generates tracked videos that maintain the identities of individual mosquitoes. Furthermore, it provides detailed locomotor activity results for both individual mosquitoes and groups of mosquitoes, including charts showing changes in distance across frames, locomotion distribution (indicating how many mosquitoes were flying simultaneously across different frames), frames during which mosquitoes were actively moving, the active-to-inactive mosquito ratio, and heat maps showing spatial movement patterns. This study further employs LocoTrackAI to analyze the locomotor activity of dengue-infected \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes using continuous video recordings. Plexiglass cages measuring 25 cm x 25 cm x 40 cm were used to facilitate the mosquitoes' natural movement, enabling detailed insights into their behavior through the tracking and analysis capabilities of LocoTrackAI.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003eThis study adopted a systematic methodology to monitor locomotor activity in \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes. The process began with mosquito rearing under controlled environmental conditions, followed by infecting mosquitoes with dengue virus to study infection-induced behavioral changes. A custom experimental setup was then designed to facilitate precise data acquisition through video recordings. Subsequent steps involved the preparation of training, validation, and test datasets to develop and evaluate the Convolutional Neural Network (CNN) model and LocoTrackAI. The model\u0026apos;s training was performed using training data and then validated using validation data. Test data was subsequently used to assess the performance of LocoTrackAI. Detailed locomotor activity analysis was performed to examine key parameters. Finally, the results were interpreted to gain insights into mosquito behavior under the influence of dengue infection, laying the groundwork for future research advancements. These steps and selected parameters, along with a summary of the results, are illustrated in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\n\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eMosquito rearing and dengue infections\u003c/h2\u003e\n \u003cp\u003eAll experimental procedures were conducted under biosafety level 3 (BSL-3) conditions within the highly secure insectary facilities at the CSIRO Australian Centre of Disease Preparedness (ACDP). These conditions ensured the safe handling and containment of dengue virus serotype 2 (DENV2) to prevent any risk of contamination or environmental exposure. \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes used in the study were maintained in a carefully controlled environment with a constant temperature of 27\u0026deg;C and a relative humidity of 70%, conditions that mimic the optimal climate for mosquito survival and activity. A 12-hour light/dark cycle was implemented to simulate natural circadian rhythms, further ensuring that the mosquitoes exhibited typical behavioral patterns. To maintain uniformity and reduce variability, all mosquitoes originated from the same batch of eggs and were reared under similar conditions, ensuring consistency in their behavioral traits. Adult mosquitoes were provided with unrestricted access to a 10% sucrose solution (\u003cem\u003ead libitum\u003c/em\u003e), which served as their primary energy source and helped maintain their vitality during the experiments. The DENV2 isolate ET300 (GenBank accession number EF440433) was selected for this study due to its relevance in understanding viral transmission dynamics. Before mosquito exposure, the virus was propagated in Vero cell monolayer cultures to ensure sufficient viral titers for infection. The exposure of \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes to DENV followed established protocols [\u003cspan class=\"CitationRef\"\u003e37\u003c/span\u003e]. Female mosquitoes were offered an infectious blood meal using an artificial feeding system, which involved chicken blood and skin to mimic natural feeding behaviors. For the uninfected control groups, female mosquitoes were fed with non-infectious blood to serve as a baseline for comparison in subsequent behavioral analyses. After the feeding process, female mosquitoes were monitored, and those that successfully engorged were carefully selected for experiments.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eExperimental setup:\u003c/h3\u003e\n\u003cp\u003eInfected and uninfected \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes were housed in custom-designed rectangular plexiglass cages, each measuring 25 cm x 25 cm x 40 cm. These cages were specifically designed to optimize the observation and recording process, featuring a movable solid white plexiglass wall on one side. This wall was included to minimize background interference, which is often caused by standard netting sleeves, and to provide a clean, unobstructed area for recording. This setup ensured that mosquito movements could be observed and analyzed with precision, particularly in relation to their locomotion behavior analysis. At twelve days post-blood feeding, groups of 5 infected mosquitoes or 5 uninfected mosquitoes were introduced into the cages for separate observation sessions. A single Flea3 camera, manufactured by Point Grey Research in Canada and equipped with infrared capture capabilities, was positioned to record the mosquitoes\u0026apos; movements (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). This setup allowed for accurate and detailed data collection, facilitating in-depth analysis of locomotion behavior.\u003c/p\u003e\n\u003ch3\u003eData collection:\u003c/h3\u003e\n\u003cp\u003eData collection for the study was performed twelve days post-infection (dpi) by capturing video recordings at a frame rate of 60 frames per second. The process involved three groups, each consisting of five infected mosquitoes, along with three additional groups, each containing five uninfected mosquitoes. The Flea3 camera, which comes with a native resolution of 1280 x 1024 pixels, was reconfigured to a frame size of 1040 x 1024 pixels. This adjustment was made to align the camera\u0026apos;s field of view with the dimensions of the mosquito cages and to reduce any unnecessary background from the recordings, ensuring that the focus remained solely on the mosquitoes. Each recording session was conducted independently for each group of infected and noninfected mosquitoes, with a duration of 30 minutes per session, resulting in a total recording time of 180 minutes.\u003c/p\u003e\n\u003ch3\u003eMosquito detection and locomotion tracking:\u003c/h3\u003e\n\u003cp\u003eMosquito detection and locomotion tracking were performed using LocoTrackAI, which is equipped with a convolutional neural network (CNN) and a multi-object tracking algorithm. This tool facilitated the automated analysis of video recordings, ensuring accurate tracking of individual mosquitoes as well as their overall activity within the experimental cages. Details about the CNN model architecture, the multi-object tracking algorithm, and the graphical user interface are provided in the sections below.\u003c/p\u003e\n\u003ch3\u003eModel architecture:\u003c/h3\u003e\n\u003cp\u003eLocoTrackAI utilizes YOLO11 for object detection (the latest version as of February 2025) [\u003cspan class=\"CitationRef\"\u003e38\u003c/span\u003e], Deep SORT for multi-object tracking [\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e], and advanced techniques for analyzing the locomotion of \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes. The YOLO11 architecture includes a distinct backbone and head design optimized for real-time detection tasks. The backbone integrates C3K2 blocks, which improve feature extraction by processing smaller feature maps through efficient convolutional layers while preserving essential details. This approach enhances both speed and accuracy compared to earlier YOLO versions. Additionally, the Spatial Pyramid Pooling Fast (SPFF) module aggregates features from different scales, improving the detection of small objects. The C2PSA (Cross Stage Partial with Spatial Attention) block further refines spatial focus, allowing the model to detect critical regions in complex scenes. The detection head employs multi-scale predictions to ensure accurate localization of objects across varying sizes. Deep SORT is employed for multi-object tracking, combining a Kalman filter to predict object states such as position, velocity, and acceleration, with the Hungarian algorithm to match objects efficiently. This process evaluates a cost matrix that incorporates motion consistency through Mahalanobis distance and appearance similarity via cosine distance, enabling robust tracking even during temporary occlusions or overlaps (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea).\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003eLocoTrackAI graphical user interface:\u003c/h2\u003e\n \u003cp\u003eThe tool utilizes a trained model, which can be selected through the \u0026quot;Select Model\u0026quot; button. A pre-trained model is included for immediate application, providing users with a ready-to-use option. Alternatively, users can train their own models by following the straightforward guidelines outlined in [\u003cspan class=\"CitationRef\"\u003e38\u003c/span\u003e]. This feature allows flexibility for both general use and customized applications. Key adjustable parameters available in the tool include Confidence, Intersection over Union (IoU), Frame Size, Threshold, and Skip Frame. The Confidence parameter is critical for filtering detections based on their confidence scores, with only those surpassing a user-specified threshold contributing to the final tracking output. By default, this parameter is set to 0.05, a setting that is particularly suitable for identifying small objects like mosquitoes, ensuring that lower-confidence but potentially relevant detections are included. The IoU parameter is employed during the Non-Maximum Suppression (NMS) stage of the detection process. It assesses the spatial overlap between bounding boxes to remove redundant predictions, thereby enhancing the accuracy of the detection results. The default IoU threshold is also set at 0.05, enabling a more lenient overlap evaluation. This default setting is particularly beneficial for detecting small or closely positioned objects, where a higher threshold might eliminate true-positive detections. The Frame Size parameter allows users to define the maximum input image size, ensuring flexibility across different resolutions. The default value is set to 1040, but LocoTrackAI automatically adjusts this value to the nearest multiple of 32 for compatibility with the neural network architecture. For instance, if the input frame size is 1040, the tool adjusts it to 1056 to optimize processing. The Threshold parameter determines locomotion by calculating the distance between consecutive frame positions of a mosquito in pixels. The tool uses Eq.\u0026nbsp;1 to classify movement.\u003c/p\u003e\n \u003cdiv id=\"Equa\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e$$\\:\\text{Distance}=\\sqrt{{\\left({x}_{f+1}-{x}_{f}\\right)}^{2}+{\\left({y}_{f+1}-{y}_{f}\\right)}^{2}}\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\left(1\\right)$$\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{f},\\:{y}_{f}\\:\\)\u003c/span\u003e\u003c/span\u003eand \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{f+1},\\:{y}_{f+1}\\:\\)\u003c/span\u003e\u003c/span\u003erepresent the coordinates of the mosquito\u0026apos;s position in consecutive frames. The resulting distance, measured in pixels, determines whether the movement exceeds the predefined threshold to be classified as locomotion. If the calculated distance exceeds the predefined threshold, the movement is classified as locomotion. The Skip Frame parameter enables users to analyze locomotion at intervals by skipping frames. This is particularly useful when dealing with computational constraints or when studying insects that exhibit slow or infrequent movements. For our analysis, the Skip Frame value was set to 0, meaning no frames were skipped during processing. Setting the value to 1 would skip every alternate frame after processing one frame (e.g., frames 1, 3, 5, etc.), while setting it to 2 would process every third frame (e.g., frames 1, 4, 7, etc.). This feature provides flexibility in balancing computational load and observation detail. After configuring these parameters, users can select a folder containing input videos using the \u0026quot;Select Folder\u0026quot; button. Processing begins automatically upon selection. For each video, LocoTrackAI generates a subfolder in the \u0026apos;Results\u0026apos; directory (within the input folder) named after the corresponding video. These subfolders contain detailed locomotion analysis results, including Excel files, charts illustrating changes in distance across frames, locomotion distribution (indicating how many insects were moving simultaneously across frames), frames where active movement occurred, the active-to-inactive mosquito ratio, heat maps showing spatial movement patterns, and statistics in text format. LocoTrackAI also allows users to stop processing at any time using the stop button. The tool ensures that any partially processed videos and their results are automatically saved. The graphical user interface of LocoTrackAI is shown in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eThreshold selection for locomotion detection:\u003c/h3\u003e\n\u003cp\u003eThe threshold for locomotion detection was set at 1.5 pixels, meaning any movement (distance \u0026ndash; Eq.\u0026nbsp;1) greater than this value was classified as locomotion. To determine this threshold, flight video segments of 20 seconds were selected from each sample group, with one segment taken from each of the three infected and three noninfected samples. These segments were processed to identify the average and lowest movement distances when mosquitoes were actively moving. The average movement distance was 4.717 pixels, and the lowest was 1.8 pixels. Based on the lowest recorded value, a threshold of 1.5 pixels was selected to ensure the detection of even minor locomotion.\u003c/p\u003e\n\u003ch3\u003eTraining, validation and test data:\u003c/h3\u003e\n\u003cp\u003eFor the customized training of the CNN model and LocoTrackAI performance validation, training, validation, and testing datasets were created using distinct videos to prevent data leakage. Each video, with a duration of 30 minutes, was converted into frames at a rate of 60 frames per second using the Python OpenCV library. For the training dataset, Video 1 from the infected group and Video 1 from the noninfected group were used, from which 4000 frames were randomly selected. The validation dataset was prepared using Video 2 from the infected group and Video 2 from the noninfected group, with 1000 frames randomly selected after frame conversion. As LocoTrackAI takes video sequences as input, the testing dataset was created using Video 3 from the infected group and Video 3 from the noninfected group, where four video sequences were selected, consisting of two sequences from the infected group and two from the noninfected group, with each sequence lasting 30 seconds and the videos having a frame rate of 60 frames per second, resulting in a total of 7204 frames (the extra frame occurs for each video because frame counting starts at 0, not 1). Since each frame contained five mosquitoes, a total of 36,020 mosquito positions were analyzed. This updated dataset split results in 80 percent for training, 20 percent for validation, and a detailed sequence-based testing approach, ensuring robust evaluation and avoiding data leakage.\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eModel training:\u003c/h2\u003e\n \u003cp\u003eThe CNN model underwent training for 1000 epochs; however, due to the early stopping mechanism with a patience setting of 50, training stopped at epoch 209, as no further improvement in performance was observed. The best performance was recorded at epoch 159. An epoch refers to a complete pass through the training data during the training process. For annotating the training and validation datasets, the Makesense web tool was employed to generate .txt files [\u003cspan class=\"CitationRef\"\u003e40\u003c/span\u003e].\u003c/p\u003e\n \u003ch2\u003e\u003cstrong\u003eLocoTrackAI evaluation\u003c/strong\u003e:\u003c/h2\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eA. Detection accuracy\u003c/h2\u003e\n \u003cp\u003eTo evaluate LocoTrackAI, the first step was assessing detection accuracy, which measures how many mosquitoes were correctly detected compared to the actual number present in each frame. Given that there are a total of 7204 test frames across four videos, with five mosquitoes per frame, a total of 36,020 mosquito instances are analyzed for detection accuracy. Detection accuracy was calculated as the ratio of correctly detected mosquitoes to the actual number of mosquitoes in the test frames, where a detection is considered correct if the Intersection over Union (IoU) between the ground truth bounding box and the bounding box calculated by LocoTrackAI is greater than 0.5. This metric provides an overall measure of how well LocoTrackAI detects mosquitoes. The IoU calculation was performed using Python and is presented in Eq.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n \u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e$$\\:IoU=\\frac{Area\\:of\\:intersection}{Area\\:of\\:union}$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eB. Centroid tracking accuracy:\u003c/h2\u003e\n \u003cp\u003eOnce detection accuracy was established, the evaluation proceeds to a more detailed analysis by assessing centroid tracking accuracy using the distance error between the ground truth centroid positions and the calculated centroid positions of mosquitoes. Each mosquito was considered an individual instance for evaluation. The Euclidean distance error for each mosquito was computed using Eq.\u0026nbsp;3.\u003c/p\u003e\n \u003cdiv id=\"Equb\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e$$\\:{E}_{i,f}=\\sqrt{{\\left({x}_{i,f}-{\\widehat{x}}_{i,f}\\right)}^{2}+{\\left({y}_{i,f}-{\\widehat{y}}_{i,f}\\right)}^{2}}\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\left(3\\right)$$\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{E}_{i,f}\\)\u003c/span\u003e\u003c/span\u003e represents the distance error for mosquito \u003cem\u003ei\u003c/em\u003e in frame \u003cem\u003ef\u003c/em\u003e. The coordinates \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{i,f}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{y}_{i,f}\\)\u003c/span\u003e\u003c/span\u003e denote the detected centroid position, while \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\widehat{x}}_{i,f}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\widehat{y}}_{i,f}\\)\u003c/span\u003e\u003c/span\u003e correspond to the ground truth centroid position of the mosquito in the same frame. This metric helps assess the accuracy of LocoTrackAI in tracking mosquito movements. A detection is considered accurate if the distance error for each mosquito instance is \u0026le;\u0026thinsp;1 pixel. If the error exceeds 1 pixel, the detection is classified as inaccurate, indicating potential tracking deviations. This threshold is necessary because mosquitoes are not single-pixel organisms, and their positions vary naturally, making precise centroid estimation inherently challenging. Additionally, manual ground truth annotation may introduce minor inaccuracies, further justifying the need for a reasonable error tolerance in evaluating tracking performance. By analyzing each mosquito individually across 7204 test frames, this approach provided a detailed accuracy evaluation, helping to identify inconsistencies in mosquito movement estimation and ensuring precise tracking performance. For ground truth centroid validation, AstroImageJ was used as it allows pixel-level position inspection with decimal accuracy (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eb). This enables precise verification of mosquito positions.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003eC. Post-occlusion identity tracking evaluation:\u003c/h2\u003e\n \u003cp\u003eTo assess the tool\u0026apos;s ability to maintain mosquito identities following occlusions, occlusions were detected using the Intersection over Union (IoU) metric (Eq.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e) to evaluate the overlap between bounding boxes identified by the Convolutional Neural Network (CNN) for mosquitoes within video frames. An IoU threshold of 0.25 (25%) was applied, indicating that when the IoU value between the bounding boxes of detected mosquitoes exceeded this threshold, the mosquitoes were classified as being occluded. This threshold, representing partial overlap, was selected to prevent higher thresholds from causing overlapping mosquitoes to be mistakenly perceived as a single entity before being categorized as occluded. To verify identity tracking, occlusions were manually reviewed by analyzing videos. The evaluation was conducted on all six videos (three infected and three noninfected), each lasting 30 minutes.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003eLocomotor activity analysis:\u003c/h2\u003e\n \u003cp\u003eThe locomotor activity of dengue-infected and noninfected mosquitoes was analyzed using 3 samples for each infected and noninfected, with each sample consisting of 108000 frames. The locomotor activity for each sample was calculated as the number of movements (changes in position) detected between consecutive frames. Average locomotor activity for each frame across all samples in both groups was computed, and this data was used to compare patterns between infected and noninfected mosquitoes. For visualization, smoothed GAM (Generalized Additive Model) values and distribution plots of locomotor activity were generated to provide a comprehensive understanding of differences between the groups. To compare the locomotor activity of infected and noninfected mosquitoes, statistical analysis was performed using the Mann-Whitney test, selected due to the non-normal distribution of the data. For statistical analysis, instead of analyzing locomotor activity on a frame-by-frame basis, batch-wise averages over 1,000 frames were used to reduce noise and variability, ensuring a more stable and representative analysis. As the total number of frames across all samples was 324,000 for infected and 324,000 for noninfected mosquitoes, this batching process resulted in 324 batch averages for each group. These batches were created by calculating the average locomotion values of all mosquitoes within the first 1,000 frames of each sample, followed by the next 1,000 frames, and so on. Shannon entropy was employed to quantify the spatial distribution of mosquitoes in the cage (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ec). This metric was calculated to assess how evenly mosquito positions were distributed across spatial sections, providing insight into differences in movement patterns and area utilization between infected and noninfected groups. Occupancy data, representing the number of mosquito positions in each section, were normalized into relative frequencies to account for differences in activity levels. Shannon entropy was then calculated based on these normalized values, with higher entropy indicating a more uniform distribution and lower values reflecting clustering in specific areas. This analysis was performed for both infected and noninfected groups to compare their spatial behavior. Statistical analysis was performed using the Python library SciPy.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003eComputation and programming system:\u003c/h2\u003e\n \u003cp\u003eLocoTrackAI was implemented on a computing system featuring an Intel\u0026reg; Core\u0026trade; i9-13980HX processor with 24 cores, capable of reaching up to 5.6 GHz on select performance cores. The system is equipped with 32GB DDR4-3200 RAM and an NVIDIA RTX 4080 GPU, operating on a Windows 11 Pro 64-bit environment. The tool framework for LocoTrackAI leverages Python 3.8.10 alongside essential scientific computing libraries, including NumPy 1.24.4 and SciPy 1.10.1, for numerical computations. Object detection and tracking are facilitated using Torch 1.9.1\u0026thinsp;+\u0026thinsp;cu111 and OpenCV 4.10.0.84, ensuring efficient deep learning-based analysis of mosquito locomotion.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eA. CNN model training results for validation data:\u003c/h2\u003e \u003cp\u003eThe object detection CNN model exhibited strong performance in accurately identifying mosquitoes. The validation results demonstrated high evaluation metrics, including a precision of 0.994, a recall of 0.992, and a mean average precision (mAP50) of 0.993. The patterns in precision, recall, and mAP50 values across the training epochs are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea-c, highlighting the model's consistent improvement during training.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eB. Test data results:\u003c/h2\u003e \u003cp\u003eLocoTrackAI effectively tracked the majority of mosquito positions across 4 test videos, demonstrating its strong performance and reliability. Out of a total of 36,020 mosquito positions, the tool successfully tracked 35,986 (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The detailed results for Test Video 1 are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e, while the data for the remaining test videos are included in the supplementary materials (Supplementary Figs.\u0026nbsp;1 to 3). For Test Video 1, LocoTrackAI detected 9,001 mosquito positions out of 9,005 across the frames. The average change in distance for an individual mosquito was recorded as 2.77 pixels (average of frames with and without locomotion) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). The number of frames in which an individual mosquito exhibited locomotor activity totaled 921 (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb). The cumulative distance change for all five mosquitoes across the frames was measured at 5.17 pixels (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec). Additionally, Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed illustrates the group activity in terms of the number of mosquitoes in motion across the frames. Analysis of group activity revealed that a single mosquito was moving in 666 frames, while in 353 frames, two mosquitoes were moving simultaneously (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ee). Figure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ef presents the active-to-inactive mosquito ratio across the frames, providing further insights into the overall group activity patterns. Furthermore, Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eg and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eh depict heatmaps that represent the spatial distribution of group activity. Figure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eg displays the heatmap for an individual mosquito, while Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eh shows the group heatmap for all mosquitoes, highlighting the number of frames and the percentage of time spent in specific regions of the test arena. All processed test videos are available in the supplementary data (Supplementary Videos 1 to 4).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eLocoTrackAI results for test videos\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVideo Number\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal Frames\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTotal Mosquito Positions\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMosquito ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePositions Per ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePositions Detected by LocoTrackAI\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e9005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1797\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e9005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1797\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e9005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1779\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e9005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1797\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7204\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e36020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e36020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e35986\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003ePerformance validation\u003c/h2\u003e \u003cp\u003eLocoTrackAI's performance was validated by assessing its effectiveness on 4 test videos using detection accuracy, centroid tracking accuracy, and performing post-occlusion identity tracking evaluation, which are detailed below.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eA. Detection accuracy:\u003c/h2\u003e \u003cp\u003eThe performance of LocoTrackAI on 4 test videos demonstrated exceptional accuracy in tracking mosquito positions. For most individual mosquitoes, the tool achieved an accuracy of 100%, with the lowest accuracy observed being 98.78% for Mosquito 4 in Video 3. The combined accuracy for all mosquitoes across the test videos was above 99%, with the lowest recorded combined accuracy being 99.76% for Video 3. The average accuracy across all test videos was 99.91% (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). These results emphasize the reliability and robustness of LocoTrackAI in tracking mosquitoes.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eLocoTrackAI evaluation results for detecting mosquito positions\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVideo Number\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal Frames\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMosquito ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePositions Per ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCorrectly Detected Positions by LocoTrackAI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eIncorrect/ Missing Detections\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eAccuracy Per ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eCombined Accuracy\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e99.96%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1797\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e99.78%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e99.96%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1797\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e99.78%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e99.76%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1779\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e98.78%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e99.96%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1797\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e99.78%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1801\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e100%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal/Average\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7204\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e36020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e35986\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e99.91%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eB. Centroid tracking accuracy:\u003c/h2\u003e \u003cp\u003eCentroid Tracking Accuracy was calculated using the Euclidean distance error, which accounted for errors across both the x and y axes. Figure\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e highlights the mean Euclidean distance error for each frame; for instance, for frame 1, it represents the mean Euclidean distance error of all 20 mosquito positions in that specific frame. To enhance the clarity of the data, a 50-frame moving average was applied, thereby facilitating better understanding of the chart. The overall average of the mean Euclidean distance errors across all test video frames was 0.22 pixels, highlighting the tool's exceptional performance.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eC. Post-occlusion identity tracking evaluation\u003c/h2\u003e \u003cp\u003eThe LocoTrackAI demonstrated outstanding performance in post-occlusion identity tracking, accurately reassigning identities in 193 out of 207 occlusion instances and achieving a success rate of 93.23%. These results highlight the tool's robustness in maintaining accurate identity tracking while tracking locomotor activity.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eLocomotor activity analysis\u003c/h2\u003e \u003cp\u003eThe locomotor activity of dengue-infected and noninfected mosquitoes was analyzed across 3 samples of each group, with each sample consisting of 108,000 frames and a total of 540,000 mosquito positions. The cumulative mosquito positions were 1.62\u0026nbsp;million for infected and 1.62\u0026nbsp;million for noninfected mosquitoes, amounting to 3.24\u0026nbsp;million positions in total. The detection accuracy for all samples exceeded 99%, ensuring the reliability of the results (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The analysis revealed that dengue-infected mosquitoes exhibited significantly higher locomotor activity compared to their noninfected counterparts. Among the infected samples, the highest locomotor activity was observed in sample 3, with 55,998 movements, while the lowest was recorded in sample 2, with 7,664 movements. The total locomotor activity across all 3 infected samples was 95,726 movements. In contrast, the noninfected samples showed their highest locomotor activity in sample 2, with 19,187 movements, and their lowest in sample 1, with 5,467, resulting in a total of 42,173 movements across all 3 samples.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eLocomotor activity results for infected and noninfected mosquitoes\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSample Number\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal Frames\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTotal Positions\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePositions Detected by LocoTrackAI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDetection Accuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eLocomotion\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003eInfected\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e108000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e540000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e539089\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99.83%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e55998\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e108000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e540000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e539871\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99.98%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7664\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e108000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e540000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e539611\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99.93%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e32064\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal/Average\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e324000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1620000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1618571\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99.91%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e95726\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003eNoninfected\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e108000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e540000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e539634\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99.93%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5467\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e108000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e540000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e539702\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99.94%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e19187\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e108000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e540000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e539005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99.82%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e17519\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal/Average\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e324000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1620000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1618341\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99.90%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e42173\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e presents the results for the first infected and noninfected mosquito samples, while the results for the remaining samples are provided as supplementary material (Supplementary Figs.\u0026nbsp;4 to 7). Figure\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ea illustrates the change in distance for infected mosquitoes in sample 1, which is 3.67 pixels, and Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eb depicts the number of frames with locomotion for the same sample. Figure\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ec shows the number of mosquitoes flying simultaneously across different frames for infected mosquitoes in sample 1, where no mosquitoes were moving for 59829 frames, one mosquito was moving for 41155 frames, two mosquitoes were moving for 6281 frames, three mosquitoes were moving for 659 frames, and four mosquitoes were moving for 76 frames. Figure\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ed provides the heatmap for infected mosquitoes in sample 1. For the noninfected mosquitoes, Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ee displays the change in distance, which is 1.37 pixels, while Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ef represents the frames with locomotion for noninfected mosquitoes in sample 1. Figure\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eg shows the number of mosquitoes flying simultaneously across different frames for noninfected mosquitoes in sample 1, where no mosquitoes were moving for 102771 frames, one mosquito was moving for 4991 frames, and tow mosquitoes were moving for 238 frames. Finally, Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eh presents the heatmap for noninfected mosquitoes in sample 1.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003ea illustrates the average locomotor activity of infected and noninfected mosquitoes across all frames, calculated by averaging the locomotor activity for each frame across all samples in each group. To provide a clearer representation, Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eb displays the smoothed GAM values for both groups. Figure\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003ec presents the distribution of average locomotor activity across all samples, demonstrating that infected mosquitoes exhibited higher locomotor activity compared to noninfected mosquitoes. The mean locomotor activity of infected mosquitoes was 0.30 movements, while for noninfected mosquitoes, it was 0.13 movements. Figure\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003ed further explores the locomotor activity by presenting values in the form of averages calculated over batches of 1,000 frames each. This data was used to perform a Mann-Whitney test, which yielded a p-value of 0.0009, confirming a statistically significant difference in locomotor activity between the infected and noninfected mosquitoes. These findings highlight the significant impact of dengue infection on the movement patterns of mosquitoes. The spatial distribution patterns of mosquitoes offer important insights into their behavioral tendencies and environmental interaction. Scatterplots presented in Figs.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003ee and \u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003ef show the spatial positions of infected and noninfected mosquitoes, respectively, illustrating their movement paths within the cage. To further clarify these patterns, Figs.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eg and \u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eh include heatmaps that represent the density of mosquito positions.\u003c/p\u003e \u003cp\u003eThe Shannon entropy values provide a systematic approach to analyzing these spatial patterns. Infected mosquitoes recorded a higher entropy value of 3.38, indicating a more uniform and extensive spatial distribution across the cage. This suggests that infected mosquitoes explore a broader range of locations within the cage. In contrast, the noninfected mosquitoes demonstrated a lower entropy value of 3.13, reflecting a more concentrated spatial distribution with activity focused in specific areas. This implies that noninfected mosquitoes engage in less exploration and exhibit a preference for localized regions.\u003c/p\u003e \u003cp\u003eThese results highlight the substantial effect of dengue infection on mosquito behavior. Infected mosquitoes exhibit increased spatial exploration and movement, potentially linked to behaviors such as enhanced host-seeking. Meanwhile, noninfected mosquitoes display more constrained spatial activity and concentrated movement, illustrating a clear behavioral distinction between the two groups.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe LocoTrackAI tool can facilitate a detailed analysis of the locomotor activity of \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes. Data generated from LocoTrackAI can be used to explore various aspects of locomotor behavior, such as activity patterns during different times of the day, movement distances, and group dynamics. Furthermore, given the tool's exceptional ability to analyze the complex movement behaviors of small organisms like mosquitoes, it presents significant potential for broader applications, including studies of locomotor activity in other species, such as flies, ticks, and zebrafish.\u003c/p\u003e \u003cp\u003eLocoTrackAI demonstrated a high level of detection accuracy, achieving a rate of 99.91%, along with a low centroid detection Euclidean distance error of 0.22 pixels, both validated on 36,020 frames. These results highlight the tool's precision and reliability in tracking mosquito locomotor activity. In addition to its strong performance, the tool provides a user-friendly interface that simplifies video analysis and supports batch processing without the need for constant supervision. Furthermore, LocoTrackAI includes a Skip Frame function to improve computational efficiency by analyzing frames at predefined intervals. It also incorporates a threshold-based classification method, allowing users to adjust sensitivity based on the organism's activity level or specific study requirements. These features collectively enhance the tool's precision, adaptability, and utility for a wide range of research applications.\u003c/p\u003e \u003cp\u003eLocoTrackAI has demonstrated strong performance in tracking the locomotor activity of mosquitoes from videos recorded at 60 frames per second, even in the presence of occlusions. However, its accuracy in maintaining correct identities may decrease when using videos with lower frame rates. Previous studies on animal tracking have recommended frame rates above 25 frames per second for organisms such as mice, zebrafish, and ants [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. Given the small size, fast movements, and complex locomotor patterns of mosquitoes, higher frame rates may be required compared to those used for tracking the locomotor activity of mice, zebrafish, and ants. Additionally, gaps in data points caused by background distortions or light reflections could reduce tracking performance. To address this, using a clean and uniform background, as demonstrated with the simple white wall in this study, is recommended. Increasing the size of the training dataset and employing anti-glare materials such as plexiglass can further improve performance by reducing environmental distractions and enhancing the quality of input data for locomotor activity analysis.\u003c/p\u003e \u003cp\u003eThis study highlights the significant impact of dengue infection on the locomotor activity of \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes. Dengue-infected mosquitoes demonstrated substantially higher activity levels compared to their noninfected counterparts, with more frequent movements and greater spatial exploration. These behavioral changes may enable them to cover larger areas, encounter more hosts [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e], and locate additional breeding sites [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e], which may increase their potential for virus transmission. In contrast, noninfected mosquitoes displayed lower activity levels and a more localized spatial distribution, suggesting restricted movement and energy conservation as a baseline behavioral pattern. This study also confirms previously observed increases in the locomotor activity of dengue-infected mosquitoes while addressing limitations of earlier studies, such as the use of small confinement tubes (1 cm x 7 cm) [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e] and discontinuous recordings [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. By utilizing continuous video recordings in larger plexiglass cages (25 cm x 25 cm x 40 cm), this study provides a more natural and comprehensive analysis of movement patterns. Overall, these findings emphasize how dengue infection modulates mosquito behavior in ways that may enhance their vectorial capacity, offering valuable perspectives for understanding disease dynamics and informing vector control strategies.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn conclusion, this study demonstrates the effectiveness of LocoTrackAI in analyzing the locomotor activity of Aedes aegypti mosquitoes and highlights the significant behavioral changes induced by dengue infection. LocoTrackAI achieved a high accuracy rate of 99.91% with a low centroid detection error of 0.22 pixels, validated on 36,020 frames. Its advanced features, such as batch processing, Skip Frame functionality, and adjustable thresholds, make LocoTrackAI a reliable tool for analyzing mosquito locomotor activity and a valuable resource for studying mosquito-pathogen interactions and vector control strategies.\u003c/p\u003e \u003cp\u003eThe behavioral analysis revealed that dengue-infected mosquitoes exhibited significantly higher locomotor activity, with a total of 95,726 movements across samples, compared to 42,173 movements in noninfected mosquitoes. The mean locomotor activity of infected mosquitoes was 0.30 movements, representing over 200% of the 0.15 observed in noninfected mosquitoes. Infected mosquitoes also showed broader spatial exploration, with higher entropy values (3.38 compared to 3.13 in noninfected mosquitoes), indicating a more uniform distribution across the cage. These behavioral differences suggest that infected mosquitoes have an increased capacity to locate hosts and spread the virus, enhancing their transmission potential.\u003c/p\u003e \u003cp\u003eFuture research could expand on this work in several ways. On the behavioral side, studies could explore the locomotor activity of mosquitoes infected with other arboviruses to determine whether the observed changes are specific to dengue or represent a broader trend. On the technological side, tools like LocoTrackAI could be further developed to automate the analysis of other critical behaviors, such as feeding activity and feeding duration, providing a more comprehensive understanding of mosquito-pathogen interactions. These advancements would contribute significantly to improving vector-borne disease models and developing more targeted mosquito control strategies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors acknowledge the capabilities of the CSIRO Australian Centre for Disease Preparedness (\u003cstrong\u003egrid.413322.5\u003c/strong\u003e) in undertaking this research, including infrastructure by the National Collaborative Research Infrastructure Strategy (NCRIS).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnimal ethics statement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe experiments for chicken blood collection were conducted under animal ethics guidelines with approval from ACDP Animal Ethics Committee (#ACDP22010).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWriting\u0026mdash;original draft preparation, N.J.; writing\u0026mdash;review and editing, A.B., A.J.L., and P.N.P.; visualization, N.J.; Software, N.J.; Data curation, N.J., A.J.L.; supervision, A.B., A.J.L., and P.N.P. All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding declaration:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was partly funded by CSIRO, which provided strategic funding to P.N.P. and N.J. The funders had no role in study design, data collection and analysis, decision to publish, or manuscript preparation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupplementary images, supplementary videos, videos for testing, and LocoTrackAI graphical user interface .exe file is available at: https://drive.google.com/drive/folders/1bxgp0GasgOswm0Ox7E71Yt6yCPtBKRUM?usp=sharing\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eOrganization, W.H., \u003cem\u003eGlobal technical strategy for malaria 2016-2030\u003c/em\u003e. 2015: World Health Organization.\u003c/li\u003e\n\u003cli\u003eWorld Health Organization. \u003cem\u003eVector-borne diseases\u003c/em\u003e. 2024; Available from: https://www.who.int/news-room/fact-sheets/detail/vector-borne-diseases.\u003c/li\u003e\n\u003cli\u003eWorld Health Organization. \u003cem\u003eMalaria in children under five\u003c/em\u003e. 2018; Available from: https://www.who.int/malaria/areas/high_risk_groups/children/en/.\u003c/li\u003e\n\u003cli\u003eFranklinos, L.H., et al., \u003cem\u003eThe effect of global change on mosquito-borne disease.\u003c/em\u003e The Lancet Infectious Diseases, 2019. \u003cstrong\u003e19\u003c/strong\u003e(9): p. e302-e312.\u003c/li\u003e\n\u003cli\u003eJaved, N., A. Bhatti, and P.N. Paradkar, \u003cem\u003eAdvances in Understanding Vector Behavioural Traits after Infection.\u003c/em\u003e Pathogens, 2021. \u003cstrong\u003e10\u003c/strong\u003e(11): p. 1376.\u003c/li\u003e\n\u003cli\u003ePadilha, K.P., et al., \u003cem\u003eZika infection decreases Aedes aegypti locomotor activity but does not influence egg production or viability.\u003c/em\u003e Mem\u0026oacute;rias do Instituto Oswaldo Cruz, 2018. \u003cstrong\u003e113\u003c/strong\u003e.\u003c/li\u003e\n\u003cli\u003ePetersen, M.T., et al., \u003cem\u003eThe impact of the age of first blood meal and Zika virus infection on Aedes aegypti egg production and longevity.\u003c/em\u003e PLoS One, 2018. \u003cstrong\u003e13\u003c/strong\u003e(7): p. e0200766.\u003c/li\u003e\n\u003cli\u003eStyer, L.M., M.A. Meola, and L.D. Kramer, \u003cem\u003eWest Nile Virus Infection Decreases Fecundity of Culex tarsalis Females.\u003c/em\u003e Journal of Medical Entomology, 2007. \u003cstrong\u003e44\u003c/strong\u003e(6): p. 1074-1085.\u003c/li\u003e\n\u003cli\u003eVogels, C.B.F., et al., \u003cem\u003eVirus interferes with host-seeking behaviour of mosquito.\u003c/em\u003e The Journal of experimental biology, 2017. \u003cstrong\u003e220\u003c/strong\u003e(Pt 19): p. 3598-3603.\u003c/li\u003e\n\u003cli\u003eMartin, E., et al., \u003cem\u003eDifferential responses of the mosquito Aedes albopictus from the Indian Ocean region to two chikungunya isolates.\u003c/em\u003e BMC Ecol, 2010. \u003cstrong\u003e10\u003c/strong\u003e: p. 8.\u003c/li\u003e\n\u003cli\u003eBhatti, A., et al., \u003cem\u003eEmerging Trends in Neuro Engineering and Neural Computation\u003c/em\u003e. 2017: Springer.\u003c/li\u003e\n\u003cli\u003eGaburro, J., et al., \u003cem\u003eZika virus-induced hyper excitation precedes death of mouse primary neuron.\u003c/em\u003e Virology journal, 2018. \u003cstrong\u003e15\u003c/strong\u003e(1): p. 1-13.\u003c/li\u003e\n\u003cli\u003eGaburro, J., S. Nahavandi, and A. Bhatti, \u003cem\u003eInsects Neural Model: Potential Alternate to Mammals for Electrophysiological Studies\u003c/em\u003e, in \u003cem\u003eEmerging Trends in Neuro Engineering and Neural Computation\u003c/em\u003e. 2017, Springer. p. 119-130.\u003c/li\u003e\n\u003cli\u003eCator, L.J., et al., \u003cem\u003e\u0026apos;Manipulation\u0026apos; without the parasite: altered feeding behaviour of mosquitoes is not dependent on infection with malaria parasites.\u003c/em\u003e Proceedings of the Royal Society B: Biological Sciences, 2013. \u003cstrong\u003e280\u003c/strong\u003e(1763): p. 20130711.\u003c/li\u003e\n\u003cli\u003eKoella, J.C., L. Rieu, and R.E.L. Paul, \u003cem\u003eStage-specific manipulation of a mosquito\u0026apos;s host-seeking behavior by the malaria parasite Plasmodium gallinaceum.\u003c/em\u003e Behavioral Ecology, 2002. \u003cstrong\u003e13\u003c/strong\u003e(6): p. 816-820.\u003c/li\u003e\n\u003cli\u003eBrady, O.J., et al., \u003cem\u003eRefining the global spatial limits of dengue virus transmission by evidence-based consensus.\u003c/em\u003e 2012.\u003c/li\u003e\n\u003cli\u003eTallon, A.K., et al., \u003cem\u003eDengue infection modulates locomotion and host seeking in Aedes aegypti.\u003c/em\u003e PLoS neglected tropical diseases, 2020. \u003cstrong\u003e14\u003c/strong\u003e(9): p. e0008531-e0008531.\u003c/li\u003e\n\u003cli\u003eSim, S., J.L. Ramirez, and G. Dimopoulos, \u003cem\u003eDengue virus infection of the Aedes aegypti salivary gland and chemosensory apparatus induces genes that modulate infection and blood-feeding behavior.\u003c/em\u003e PLoS Pathog, 2012. \u003cstrong\u003e8\u003c/strong\u003e(3): p. e1002631.\u003c/li\u003e\n\u003cli\u003eSylvestre, G., M. Gandini, and R. Maciel-de-Freitas, \u003cem\u003eAge-dependent effects of oral infection with dengue virus on Aedes aegypti (Diptera: Culicidae) feeding behavior, survival, oviposition success and fecundity.\u003c/em\u003e PLoS One, 2013. \u003cstrong\u003e8\u003c/strong\u003e(3): p. e59933.\u003c/li\u003e\n\u003cli\u003eMaciel-de-Freitas, R., et al., \u003cem\u003eThe Influence of Dengue Virus Serotype-2 Infection on Aedes aegypti (Diptera: Culicidae) Motivation and Avidity to Blood Feed.\u003c/em\u003e PLOS ONE, 2013. \u003cstrong\u003e8\u003c/strong\u003e(6): p. e65252.\u003c/li\u003e\n\u003cli\u003eGaburro, J., et al., \u003cem\u003eDengue virus infection changes Aedes aegypti oviposition olfactory preferences.\u003c/em\u003e Scientific Reports, 2018. \u003cstrong\u003e8\u003c/strong\u003e(1): p. 13179.\u003c/li\u003e\n\u003cli\u003eLima-Camara, T.N., et al., \u003cem\u003eDengue infection increases the locomotor activity of Aedes aegypti females.\u003c/em\u003e PloS one, 2011. \u003cstrong\u003e6\u003c/strong\u003e(3): p. e17690-e17690.\u003c/li\u003e\n\u003cli\u003eGaburro, J., et al., \u003cem\u003eNeurotropism and behavioral changes associated with Zika infection in the vector Aedes aegypti.\u003c/em\u003e Emerg Microbes Infect, 2018. \u003cstrong\u003e7\u003c/strong\u003e(1): p. 68.\u003c/li\u003e\n\u003cli\u003eAlShamsi, M., et al., \u003cem\u003eArtificial intelligence and blockchain for transparency in governance\u003c/em\u003e, in \u003cem\u003eArtificial Intelligence for Sustainable Development: Theory, Practice and Future Applications\u003c/em\u003e. 2021, Springer. p. 219-230.\u003c/li\u003e\n\u003cli\u003eMatthews, G., et al., \u003cem\u003eEvolution and revolution: Personality research for the coming world of robots, artificial intelligence, and autonomous systems.\u003c/em\u003e Personality and individual differences, 2021. \u003cstrong\u003e169\u003c/strong\u003e: p. 109969.\u003c/li\u003e\n\u003cli\u003eNawaz, M.S., et al., \u003cem\u003eUsing artificial intelligence techniques for COVID-19 genome analysis.\u003c/em\u003e Applied Intelligence, 2021: p. 1-18.\u003c/li\u003e\n\u003cli\u003eYang, S.J., et al., \u003cem\u003eHuman-centered artificial intelligence in education: Seeing the invisible through the visible.\u003c/em\u003e Computers and Education: Artificial Intelligence, 2021. \u003cstrong\u003e2\u003c/strong\u003e: p. 100008.\u003c/li\u003e\n\u003cli\u003eShabbir, J. and T. Anwer, \u003cem\u003eArtificial intelligence and its role in near future.\u003c/em\u003e arXiv preprint arXiv:1804.01396, 2018.\u003c/li\u003e\n\u003cli\u003eJaved, N., et al., \u003cem\u003eEggCountAI: a convolutional neural network-based software for counting of Aedes aegypti mosquito eggs.\u003c/em\u003e Parasites \u0026amp; Vectors, 2023. \u003cstrong\u003e16\u003c/strong\u003e(1): p. 341.\u003c/li\u003e\n\u003cli\u003eJaved, N., et al., \u003cem\u003eFlightTrackAI: a robust convolutional neural network-based tool for tracking the flight behaviour of Aedes aegypti mosquitoes.\u003c/em\u003e Royal Society Open Science, 2024. \u003cstrong\u003e11\u003c/strong\u003e(10): p. 240923.\u003c/li\u003e\n\u003cli\u003eJaved, N., et al., \u003cem\u003eLarvaeCountAI: a robust convolutional neural network-based tool for accurately counting the larvae of Culex annulirostris mosquitoes.\u003c/em\u003e Acta Tropica, 2024. \u003cstrong\u003e260\u003c/strong\u003e: p. 107468.\u003c/li\u003e\n\u003cli\u003eJaved, N., et al., \u003cem\u003eFlight traits of dengue-infected Aedes aegypti mosquitoes.\u003c/em\u003e Computers in Biology and Medicine, 2024. \u003cstrong\u003e171\u003c/strong\u003e: p. 108178.\u003c/li\u003e\n\u003cli\u003eJaved, N., P.N. Paradkar, and A. Bhatti, \u003cem\u003eFlight behaviour monitoring and quantification of aedes aegypti using convolution neural network.\u003c/em\u003e PloS one, 2023. \u003cstrong\u003e18\u003c/strong\u003e(7): p. e0284819.\u003c/li\u003e\n\u003cli\u003eJaved, N., P.N. Paradkar, and A. Bhatti, \u003cem\u003eAn overview of technologies available to monitor behaviours of mosquitoes.\u003c/em\u003e Acta Tropica, 2024. \u003cstrong\u003e258\u003c/strong\u003e: p. 107347.\u003c/li\u003e\n\u003cli\u003eSharifrazi, D., et al., \u003cem\u003eAutomated detection of Zika and dengue in Aedes aegypti using neural spiking analysis: A machine learning approach.\u003c/em\u003e Biomedical Signal Processing and Control, 2024. \u003cstrong\u003e96\u003c/strong\u003e: p. 106594.\u003c/li\u003e\n\u003cli\u003eJoshi, A. and C. Miller, \u003cem\u003eReview of machine learning techniques for mosquito control in urban environments.\u003c/em\u003e Ecological Informatics, 2021. \u003cstrong\u003e61\u003c/strong\u003e: p. 101241.\u003c/li\u003e\n\u003cli\u003eDuchemin, J.-B., et al., \u003cem\u003eZika vector transmission risk in temperate Australia: a vector competence study.\u003c/em\u003e Virology journal, 2017. \u003cstrong\u003e14\u003c/strong\u003e(1): p. 1-10.\u003c/li\u003e\n\u003cli\u003eJocher, G., A. Chaurasia, and J. Qiu. \u003cem\u003eUltralytics YOLO\u003c/em\u003e. 2023; Available from: https://github.com/ultralytics/ultralytics.\u003c/li\u003e\n\u003cli\u003eWojke, N., A. Bewley, and D. Paulus. \u003cem\u003eSimple online and realtime tracking with a deep association metric\u003c/em\u003e. in \u003cem\u003e2017 IEEE international conference on image processing (ICIP)\u003c/em\u003e. 2017. IEEE.\u003c/li\u003e\n\u003cli\u003ePiotr Skalski. \u003cem\u003eMake Sense\u003c/em\u003e. 2019; Available from: https://www.makesense.ai.\u003c/li\u003e\n\u003cli\u003eP\u0026eacute;rez-Escudero, A., et al., \u003cem\u003eidTracker: tracking individuals in a group by automatic identification of unmarked animals.\u003c/em\u003e Nature Methods, 2014. \u003cstrong\u003e11\u003c/strong\u003e(7): p. 743-748.\u003c/li\u003e\n\u003cli\u003eRomero-Ferrero, F., et al., \u003cem\u003eidtracker.ai: tracking all individuals in small or large collectives of unmarked animals.\u003c/em\u003e Nature Methods, 2019. \u003cstrong\u003e16\u003c/strong\u003e(2): p. 179-182.\u003c/li\u003e\n\u003cli\u003eEvans, O., et al., \u003cem\u003eIncreased locomotor activity and metabolism of Aedes aegypti infected with a life-shortening strain of Wolbachia pipientis.\u003c/em\u003e Journal of Experimental Biology, 2009. \u003cstrong\u003e212\u003c/strong\u003e(10): p. 1436-1441.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Deakin University","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Locomotor activity, dengue-infected mosquitoes, artificial intelligence, post-occlusion, spatial analysis, arboviruses","lastPublishedDoi":"10.21203/rs.3.rs-6067469/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6067469/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMonitoring the locomotor activity of mosquitoes is vital for understanding their behavioral patterns and role in disease transmission. Studies investigating the locomotor activity of dengue-infected mosquitoes have faced several limitations, including confining mosquitoes within small tubes that restrict natural movement, discontinuous recordings that fail to provide detailed activity patterns, and the lack of open-source tools to effectively monitor mosquito locomotor activity. Here, we present LocoTrackAI, a robust artificial intelligence-based tool that leverages a convolutional neural network (CNN) and a multi-object tracking algorithm to comprehensively analyze the locomotor activity of \u003cem\u003eAedes aegypti\u003c/em\u003e mosquitoes from videos recorded using standard laboratory cages. LocoTrackAI automatically processes video datasets, tracks individual mosquito identities, and provides detailed locomotion results for both individual mosquitoes and group activity, including spatial distributions, movement patterns, heat maps, and activity ratios. The tool features a Skip Frame function to improve computational efficiency, adjustable movement thresholds for customized sensitivity, and a user-friendly interface that supports unsupervised batch processing, ensuring accuracy and flexibility for diverse research applications. The LocoTrackAI achieved 99.91% accuracy with a low centroid detection error of 0.22 pixels across 36,020 frames and demonstrated a 93.23% success rate in reassigning identities during 207 post-occlusion instances. Using LocoTrackAI, we analyzed the locomotor activity of dengue-infected and noninfected mosquitoes across 3.24\u0026nbsp;million recorded mosquito positions. Results revealed that infected mosquitoes exhibited significantly higher locomotor activity (p\u0026thinsp;=\u0026thinsp;0.0009), with 95,726 movements (0.30 mean locomotor activity) compared to 42,173 movements (0.13 mean locomotor activity) in noninfected mosquitoes, representing more than 200% of the activity of noninfected mosquitoes. Additionally, spatial analysis indicated a more extensive and uniform distribution for infected mosquitoes, with entropy values of 3.38 for infected and 3.13 for noninfected mosquitoes. These findings suggest that dengue infection increases locomotor activity and spatial exploration, potentially enhancing the mosquitoes' capacity to locate hosts and spread the virus. Future studies could expand on this work by investigating the locomotor effects of other arboviruses and further developing tools to automate the analysis of feeding and other critical behaviors.\u003c/p\u003e","manuscriptTitle":"LocoTrackAI: advanced convolutional neural network-based tool for monitoring locomotor activity in dengue-infected Aedes aegypti mosquitoes","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-24 11:00:18","doi":"10.21203/rs.3.rs-6067469/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"15d4eea7-1c6f-4b1b-9500-d351252e1e55","owner":[],"postedDate":"February 24th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":44598319,"name":"Infectious Diseases"},{"id":44598320,"name":"Computational Biology"},{"id":44598321,"name":"Behavioral Ecology"},{"id":44598322,"name":"Artificial Intelligence and Machine Learning"},{"id":44598323,"name":"Entomology"}],"tags":[],"updatedAt":"2025-02-24T11:00:18+00:00","versionOfRecord":[],"versionCreatedAt":"2025-02-24 11:00:18","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6067469","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6067469","identity":"rs-6067469","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}