Integrating Microfluidics and Deep Learning to Investigate Entomopathogenic Nematode Responses to Host Cues

Entomopathogenic nematodes (EPNs) are effective biocontrol agents, reducing pesticide impact on health and the environment. Understanding their physiology and ethology is crucial for optimizing their application. This study offers innovative insights about Steinernema carpocapsae EPN behavior, contributing to the interdisciplinary field of engineering, biology, and entomology. The proposed hybrid approach combines microfluidics, deep learning, and optical flow. A Convolutional Neural Network (CNN) model discerned EPNs in the presence of stimuli within a specially designed microfluidic arena, highlighting motor behavior differences. At the video level, the CNN model accurately discriminated a context characterized by host-borne cues, achieving an overall accuracy of 0.938, precision of 1, and f1-score of 0.933. Integrating optical flow analysis unveiled a significant difference in motor activity, adding novel information on their dynamic responses. EPNs showed an increased activity in stimulus presence. This comprehensive approach advances our capability to detect and comprehend the responses of EPN to host stimuli for more precise and targeted biocontrol strategies.