Comparative Assessment of Manual Segmentation of Cerebral Infarction Lesions in Experimental Animals Based on Magnetic Resonance Imaging Using Artificial Intelligence.

Abstract

The aim of this study was to evaluate the quality of manual segmentation of cerebral infarction lesions in experimental animals with modeled brain infarct based on magnetic resonance imaging compared to an automated artificial intelligence approach. For automated infarct segmentation, an artificial intelligence system with the Swin-UNETR architecture was used, while manual segmentation was performed by four independent researchers. It was shown that manual segmentation exhibits significant variability, especially when small brain infarct lesions are analyzed. The obtained data emphasize the need for standardizing methods and applying automated systems to improve the accuracy and reproducibility of the results.