| Abstract |
Variation in Drosophila compound eye size is studied across research fields, from evolutionary biology to biomedical studies, requiring the collection of large datasets to ensure robust statistical analyses. To address this, we present EyeHex, a Matlab-based tool for automatic segmentation of fruit fly compound eyes from brightfield and scanning electron microscopy (SEM) images. EyeHex features two integrated modules: the first uses machine learning to generate probability maps of the eye and ommatidia locations, while the second, a hard-coded module, leverages the hexagonal organization of the compound eye to map individual ommatidia. This iterative segmentation process, which adds one ommatidium at a time based on registered neighbors, ensures robustness to local perturbations. EyeHex also includes an analysis tool that calculates key metrics of the eye, such as ommatidia count and diameter distribution across the eye. With minimal user input for training and application, EyeHex achieves exceptional accuracy (>99.6% compared to manual counts on SEM images) and adapts to different fly strains, species, and image types. EyeHex offers a cost-effective, rapid, and flexible pipeline for extracting detailed statistical data on Drosophila compound eye variation, making it a valuable resource for high-throughput studies.
|
