Automated detection of amyloid plaques (AP) in post mortem brain sections of patients with Alzheimer disease (AD) or in mouse models of the disease is a major issue to improve quantitative, standardized and accurate assessment of neuropathological lesions as well as of their modulation by treatment. We propose a new segmentation method to automatically detect amyloid plaques in Congo Red stained sections based on adaptive thresholds and a dedicated amyloid plaque/tissue modelling. A set of histological sections focusing on anatomical structures was used to validate the method in comparison to expert segmentation.

Abstract. Automated detection of amyloid plaques (AP) in post mortem brain sections of patients with Alzheimer disease (AD) or in mouse models of the disease is a major issue to improve quantitative, standardized and accurate assessment of neuropathological lesions as well as of their modulation by treatment. We propose a new segmentation method to automatically detect amyloid plaques in Congo Red stained sections based on adaptive thresholds and a dedicated amyloid plaque/tissue modelling. A set of histological sections focusing on anatomical structures was used to validate the method in comparison to expert segmentation. Original information concerning global amyloid load have been derived from 6 mouse brains which opens new perspectives for the extensive analysis of such a data in 3-D and the possibility to integrate in vivo-post mortem information for diagnosis purposes.