Authors

Discussion

At least 2 billion people worldwide are carrying a parasitic helminth infection resulting in chronic illness, and morbidity, equating to > 55 million disability adjusted life years (DALYs). The domestic livestock industry is also affected, with >£110 million lost in revenue per annum and £360 million spent on preventative anthelmintics in the EU alone. Current chemotherapeutic options for helminth infections in both human and veterinary medicine are limited. With building concern that helminths may be losing sensitivity or building resistance to these treatments, new cost effective, strategic routes for drug discovery are needed.

The Roboworm platform is a high-throughput, high content screening system. Through the Welsh government funded Life Sciences Bridging Fund, this system has been adopted to assess whether it is possible to build an image analysis model based on the phenotype of newly excysted juvenile (NEJ) stage of helminth parasite Fasciola hepatica in collaboration with commercial partner Informatics Unlimited.

In order to capture a wide variety of phenotypes expressed by drug treated NEJs, 25 parasites per well were cultured in 10uM of model compounds (Artesunate, Nitroxynil, Clorsulon, Closantel, Auranofin) along with the current Fascioliasis treatment option, Triclabendazole, and negative control DMSO. Following 72 hours culture, NEJs were imaged at a 10x magnification (9 tiles per well). Following image capture, NEJs were segmented, individually scored as affected, unaffected or ignored, with additional subcategory considerations of image clarity, contrast and segmentation being noted.

Individual field-of-view (FOV) images were first stitched together into well images using a novel algorithm that addresses uneven illumination across individual FOV images. The obtained well images were then processed using a sequence of morphological, thresholding and segmentation operations to extract individual parasite images. The resulting objects were then assessed for their quality and a set of 75 histogram-based, texture and morphological descriptors were calculated for individual parasite images. They were next randomly split into a training and tests sets which were used to build and validate a Bayesian classification model that distinguishes between healthy and treated parasites.

Current results and planned improvements of this automated anthelmintic screening system for F. hepatica NEJs will be presented.