Application of Multispectral Imaging Combined with Data Analytics to Detect Carcass Condemnations during Poultry Processing

Abstract

A concern for consumers when buying food is the safety of their food and the risk of foodborne illnesses due to contaminated products. The poultry industry deals with known foodborne pathogens, Salmonella spp. and Campylobacter spp., that can be found in the gut biome of chicken that are being processed daily. To reduce likelihood contamination of pathogens spreading throughout processing plants, every chicken is visually inspected by USDA-FSIS inspectors or trained processing plant workers when processing whole chicken carcasses. Processing lines move at 140-175 birds per minute making it a difficult to inspect each bird properly for any carcass condemnations such as septicemia-toxemia (sep-tox) and fecal contamination, being a zero tolerance in all food processing plants enforced by USDA-FSIS ruling. With a real-time fluorescent spectral imaging system that can identify carcass condemnations, processing would be easier to target which carcasses will need to be removed for reprocessing or discarded if deemed unacceptable for consumption. With technology advancing daily, improving the inspection system on the processing line is beneficial for plant workers and the safety of consumers’ health. After configuring promising camera parameters with two exposure times for fluorescence of photosensitive cells with camera range to explore the advantages of a spectral imaging system, known as the CSI-D+ system, with the detection of sep-tox birds deemed by trained processing workers and fecal contamination on chicken carcasses. Images produced were used for image analysis for classification against carcasses that were condemned. Sep-tox birds were 100% classified correctly against normal carcasses when fluoresced with LED light. Further analysis revealed variations of sep-tox can be identified with an unsupervised image analysis system. Fecal contamination and digesta were identified from the ceca (46%), colon (39%), proventriculus (62%), and small intestine (59.50%) with small and large applications onto the breast of chicken carcasses with variation of small and large application dots. Additionally, deionized water rinsed fecal contamination carcasses were classified with a 97.80% accuracy against chicken carcasses samples inoculated with Salmonella Typhimurium. Rinsed inoculated samples were swabbed and tested with a Salmonella PCR assay for further analysis. The presence of Salmonella after rinsing was 97.14% for samples that were initially negative before inoculation. These results indicate that spectral imaging can reduce the spread of contamination by decreasing food safety risks at poultry processing plants through providing a resource imaging system for plant workers in comparison to visual inspection.