Towards automated detection of milk spot livers by diffuse reflectance spectroscopy Original Research Article

Swine ascariasis results in milk spot livers which are manually removed from the slaughter line. In this paper, we propose and objectively assess a novel automated method for detection of milk spot livers based on digital color imaging, visible and near infrared spectroscopy and multispectral measurements. A commercial digital color camera based system provided insufficient information for reliable detection of milk spot livers with respect to different focal hepatic changes. The obtained average classification sensitivity and specificity was low, i.e. 77%. The spectroscopic and multispectral systems in the VIS and NIR spectral range performed significantly better. The obtained average classification sensitivity and specificity of milk spots was 98%. A practical, multispectral approach utilizing six LED or laser diodes with the emission wavelengths centered around 460, 600, 850, 1070, 1150 and 1350 nm was found sufficient for reliable classification of milk spot livers.