AUTOMATIC IDENTIFICATION AND SEPARATION OF SOMATIC EMBRYOS IN VITRO

The heterogeneity of somatic embryo maturity and quality typically found in embryogenic cultures has impeded the commercialization of this propagation technique. This variability neccesitates identifying viable embryos in culture and separating them from the immature embryos and abberant structures prior to ex vitro processing. In this work a fluidic, in vitro embryo monitor and harvester was developed for liquid based, bioreactor culture systems. A machine vision system was used to rate the viability of embryos as they were pumped through a 3mm, square glass conduit. Embryo morphologies were ranked with 17 geometric features processed by a neural network. Embryos which ranked as harvestable by the network were ejected from a gap in the flow conduit by a precisely timed injection of medium from a control nozzle. The position and velocities of embryos between the imaged section of the conduit and the harvest gap were monitored by an object tracker employing 30 LED/photo-diode pairs mounted on 2.54mm centers along the conduit. Objects not harvested were routed back to the bioreactor. The harvester successfully removed 28% of harvestable embryos from a test population. The homogeneity of the harvested population was 88% compared to the test population homogeneity of 56%.