Automated assessment of footpad dermatitis in broiler chickens at the slaughter-line: Evaluation and correspondence with human expert scores

Footpad dermatitis is increasingly used as an indicator of decreased broiler welfare, and automation of dermatitis monitoring potentially reduces the effort needed to monitor commercial flocks. In this study we evaluated a prototype system for the automatic assessment of footpad dermatitis in broiler chickens by comparing the automatic assessment with a human expert assessment. The expert aimed at selecting 2 times (different period) 20 broilers per footpad dermatitis category (5 categories in total), from 2 different flocks of 38-d-old broilers on an experimental farm. Two days later these broilers were transported to the slaughterhouse, where footpad dermatitis was assessed by the automatic system. Subsequently the footpads were reassessed by the same expert that had selected the birds. Automatic scores were only weakly correlated with scores given by the expert on-farm (r = 0.54) and at the slaughterhouse (r = 0.59). Manual evaluation of the photographs on which the automatic system based its scores revealed several errors. For 41.1% of the birds, the automatic system assessed only one of the footpads, whereas for 15.2% neither footpad was assessed. For 49.4% of the birds, scores were based on partially incorrectly identified areas. When data from such incomplete and obviously incorrect assessments were discarded, stronger correlations between automatic and expert scores were found (r = 0.68 and r = 0.74 for expert scores given on-farm and at-slaughter, respectively). Footpads that were missed by the automatic system were more likely to receive a high expert score at slaughter (P = 0.02). However, average flock scores did not differ greatly between automatic and expert scores. The prototype system for automatic dermatitis assessment needs to be improved on several points if it is to replace expert assessment of footpad dermatitis.