Long-term imaging of living biological specimens is important to infer behaviorial trends and correlate neural structure with behavior. Such study is plagued by the field of view limitation in standard optical microscopes, as the motile specimen would frequently move out of view. A novel microscope, called the adaptive scanning optical microscope (ASOM), has recently been proposed to address this limitation. Through high speed post-objective scanning with a steering mirror, and compensation for optical aberrations with a MEMS deformable mirror, simultaneous imaging and tracking of multiple Caenorhabditis elegans worms has been demonstrated. This article presents the image processing algorithm for tracking multiple worms. Since the steering mirror has to move based on the predicted worm motion, image processing and stable steering mirror motion need to be executed at higher than the composite mosaic video frame rate (in contrast to existing works on image-based worm tracking, which are predominantly based on post-processing). Particular care is placed on disambiguating the worms when they overlap, collide, or entangle, where the worm tracking algorithm may fail. Results from both real time and simulated tracking are presented.
International Journal on Optomechatronics, 4(1), March, 2010, pp. 1-21.