Разработка нейросетевого алгоритма для определения принадлежности моногеней к отряду Dactylogyridea

Aquaculture as one of the fastest growing food-producing sectors has given rise to numerous fish farms which often lack the laboratory facilities and qualified professionals to accurately diagnose disease and provide appropriate treatments. Monogeneans are parasitic worms that belong to the phylum Platyhelminthes. Some of them can cause mass mortality of fish in both natural water bodies and aquaculture settings. This paper presents a neural network algorithm designed to accurately identify monogeneans of the order Dactylogyridea using digital photographs taken through the ocular lens of the light microscope with a smartphone camera. The lowered equipment requirements make this method suitable for use on fish farms. We frame the identification of Dactylogyridea as a binary classification problem and train the VGG‑16 convolutional neural network to classify images of these monogeneans. To enhance our dataset, we apply data augmentation techniques that artificially increase the number of training examples and simulate variations in microscope illumination levels, image under- / overexposure or parasite discolouration on a slide. Our recognition algorithm achieves a classification accuracy of 98.8 % for the elements of both testing and validation sets. The results obtained in this study are of practical value since many species of the order Dactylogyridea can cause lethal diseases in fish. This method can improve diagnostics, treatment and disease prevention in aquaculture. Additionally, its simplicity is particularly advantageous for novice specialists