AI-based artery segmentation and plaque quantification in mouse models is the subject of a new publication that appeared on 23 April in the renowned journal ‘Scientific Reports’. In this study, a multi-stage procedure was described with which Oil Red O-stained histological sectional images of the aortic root can be analysed automatically. Atherosclerotic plaques are lipid-rich deposits in the arteries whose growth and composition are of particular interest for cardiovascular research.
Due to the large number of possible secondary diseases, atherosclerosis is the main cause of death in industrialised nations. A major challenge in the research of therapeutic measures is the large amount of image data, which is currently mainly analysed manually. Thanks to AI-supported methods, these processes can now be carried out more objectively and faster than before, without having to perform time-consuming manual segmentation.
The concept developed at Fraunhofer IMTE since 2022 combines several machine learning approaches (see Figure 1). First, a region of interest of the arterial structure is determined, which makes it possible to process the fine details of the vessels in the highest possible resolution. Subsequently, so-called ensembles are used, which combine different local optima, thereby reducing false segmentations. Finally, an unsupervised method is used to determine the deposits in the arteries, which recognises these pathological changes based on the characteristic colour patterns. A special method for transferring colour spaces also minimises differences in staining and image acquisition.
This pipeline therefore utilises both supervised and unsupervised AI methods to ensure the most complete and robust detection possible. The method is characterised above all by its rapid applicability to very large data sets and thus considerably simplifies laboratory workflows.
