EyeHex toolbox for complete segmentation of ommatidia in fruit fly eyes

Variation in Drosophila compound eye size is studied in various research contexts ranging from evolutionary to biomedical studies. Studying this variation requires the collection of large datasets required for the statistical power of the subsequent analyses. EyeHex is a new tool that automatically and fully segments fruit fly compound eyes from both brightfield and scanning electron microscopy images. The system comprises two modules. The first module uses machine-learning to extract the compound eye and ommatidia regions from different types of images. The second module is hard-coded and accurately determines the number and location of individual ommatidia using the hexagonal arrangement of the compound eye. EyeHex is equipped with a manual correction tool that enables control over segmentation accuracy, particularly at the edges of the eyes where ommatidia detection is more challenging due to highly slanted imaging angles and interfering ocular hairs. EyeHex is accompanied by an analysis tool to quantify ommatidia count, position, altitude, density across the eye, and arrangement along the anterior/posterior. Training and applying EyeHex model for ommatidia segmentation requires limited user inputs. It achieves a high level of accuracy (>99% when compared to manual counting) from both Scanning Electron Microscope and multi-focus brightfield images of two Drosophila melanogaster strains. This demonstrates that EyeHex can be integrated flexibly in a cheap and fast pipeline to extract statistics on variations of Drosophila compound eyes.