Experimental study of filtration activity in Ditrupa arietina (Annelida Polychaeta) using an automated image analysis system

An automated image analysis system designed to assess the activity of benth ic organisms is described. This system is used to study filtration in the s erpulid polychaete Ditrupa arietina. The video sensor of the system is comp osed of a black and white charged coupled device, a microprocessor, a memor y and an interface board. It is driven by real-time routines, which are dow nloaded in permanent memory prior to each experiment. These routines contro l picture acquisition frequency and compute the differences in grey levels between the image recorded at a given time and a reference image (correspon ding to no filtration in the case of D. arietina). These differences are us ed to detect numerical objects, which are stored in the memory board. At th e end of each experiment, these objects are uploaded to a microcomputer whe re they are analysed using a second set of programs. This procedure involve s several parameters, namely. the minimal object surface which is indicativ e of a real difference between two images (minimal surface), the research a rea (search radius), the research sites, and the definition of a set of con ditions relating differences between images with a true filtering activity by D. arietina. These parameters were determined by comparing the results o btained using: 1) the automated system, and 2) classical frame by frame vid eotape analysis. They were then validated on several batches of worms by co mparing total filtering durations per worm measured using the as previously calibrated automated system with those obtained using classical video obse rvations. This step showed that the automated system is suitable for studyi ng filtering activity in D. arietina. Our first results show that inter-ind ividual variability is high, which has important consequences on experiment al plans designed to assess the effects of environmental factors on filtrat ion. (C) 2000 Ifremer/CNRS/IRD/Editions scientifiques et medicales Elsevier SAS.