Morphological Analysis and Classification of Vector Mosquitoes using Novel Deep Learning Model

Introduction: Vector-borne diseases, particularly those transmitted by mosquitoes, pose persistent global public health challenges. Accurate and timely identification of mosquito species is critical for controlling diseases like malaria, dengue, chikungunya, yellow fever, and Zika, each linked to specific mosquito vectors. Objective: This study proposes a novel approach for vector mosquito classification through transfer learning with an ensemble model that integrates three robust architectures: VGG-16, DenseNet-121, and ResNet-50. The strategy aims to improve detection accuracy for mosquito species associated with these vector-borne diseases. Methodology: The ensemble model leverages the complementary strengths of VGG-16, DenseNet-121, and ResNet-50 to enhance classification accuracy, targeting mosquito species involved in disease transmission. The model can adapt quickly to specific identification needs by employing a transfer learning framework, thus aiding disease prevention efforts. Results: The ensemble model achieves high classification performance, with precision, recall, and F1-score at 98% each and accuracy at 98.33%. Additional metrics, including AUC, Kappa, and Matthews Correlation Coefficient (MCC), underscore the model's robustness, surpassing previous efforts in this field. Conclusion: This study provides valuable insights into advancing vector mosquito classification systems, highlighting their potential impact on disease control measures, ecological surveillance, and broader public health initiatives. The findings emphasize the importance of enhanced classification techniques in supporting effective, data-driven responses to vector-borne disease challenges.