Enhancing deep learning for parasite detection: An integrated review of data augmentation methods for fluorescence microscopy images

Fluorescence microscopy represents an important part of parasitological research, as it visualizes parasite morphology and different life stages of the parasites in the host and provides details on interactions with host cells.The study of parasites using fluorescence imaging is a critical area of research, particularly for diagnosing and understanding diseases.The availability of large, annotated fluorescence image datasets of parasites is limited, necessitating the use of data augmentation techniques to enhance the volume and variability of the available data.This paper surveys the different data augmentation techniques that could be applied to fluorescence microscopy images of parasites in small datasets.First, the investigation will involve the traditional methods: geometric transformations, cropping, rotation, and flipping.Such techniques are foundational in nature, avoiding overfitting and increasing dataset diversity.The set of sophisticated techniques and tools in this area include GAN, synthetic data generation, colour space adjustments, mosaic augmentation, noise injection, etc., enabling the creation of far more realistic and diverse training samples.We also discuss the challenges in detecting waterborne intestinal parasites, such as low parasite prevalence, matrix interference, morphological variability, and limited availability of high-quality reference images.Addressing these challenges through effective data augmentation can significantly enhance the performance of machine learning models for tasks such as parasite classification, segmentation, and detection.Despite the advancements, several key gaps remain, including the need for larger annotated datasets, improved model generalizability, and enhanced computational efficiency.This survey aims to provide a comprehensive overview of data augmentation strategies to advance the field of parasitology, ultimately leading to improved diagnostic capabilities and more efficient workflows in clinical and research settings.