REU Student Project Abstract

The ability to monitor animals within a study gives the observers the ability to monitor and gather data on the animals without influencing their behaviors. A prominent condition typically found in infants that consume formula causes gut inflammation, creating a need to study outcomes and interventions. An experimental study was conducted to assess how the amino acid TBCD being added to the inflammatory agent DSS within infant formula can help aid in the curing of the stomach inflammation. Collective behavior of the piglets; the videos were moved to a database where the observers save them to label the behaviors. Within the DSS study data set, videos are being analyzed through Behavioral Observation Research Interactive Software (BORIS), an open-source behavioral labeling tool. The dataset being utilized was developed from a study within the Piglet Nutrition & Cognition Laboratory at the University of Illinois at Urbana-Champaign, where piglets were used within nutritional studies as a model for human infants. Pre-weaned, individually house piglets were a part of three different groups; the control, the DSS, which is the inflammatory agent, and the DSS+TBCD, which is a nutritional supplement aimed to act against and solve the inflammation. Observers using an ethogram with the various behaviors split into two categories, active and inactive, was used within the tool. The ethogram was developed through 72 hours of a continuously labeled dataset for 3 piglets; the information collected was used to list all the behavior labels and their definitions. The labeling methods used were continuous and instantaneous. The first five minutes of video was labeled using the continuous sampling method for only active behaviors. Starting at the five-minute mark, they switch over to instantaneous labeling and begin marking all inactive behaviors at every five-minute mark. An observer would watch a 48-minute video and label 45 minutes to keep up with the consistent five-minute (7500 frames) labeling increments. Over 300 videos were labeled and to be entered into the dataset to teach a computer program to detect sickness symptoms within piglets.