Robust and accurate recognition of veins in fruit fly wings

Manual inspection of the morphological differences in the veins of wings in fruit fly mutants is useful for uncovering the relevant genes associated with fundamental biological functions. However, manual morphological inspection is not feasible when processing numerous mutants that were systematically created using fruit fly genome information. Another difficulty is how to process mutant wings that violate the typical vein pattern in the wild-type fruit fly. We require an automatic image recognition system to quantify morphological differences robustly and accurately, but this is a difficult task because exact quantification of significant differences from the typical pattern is highly nontrivial. We propose a robust and accurate image recognition technique that can analyze the veins of fruit fly wings robustly and accurately using a hierarchical model matching algorithm. The results of an experiment with 1000 images demonstrated the feasibility of our technique.