Multiscale Image Registration for Stroke Lesion Analysis in Rat Brain MR and Atlas Images

With the rapid development of medical imaging technology, modern imaging techniques reveal distinct image information. Image registration can integrate multimodal images to present the corresponding physiological tissues, which plays an indispensable role in clinical treatment and analysis. This paper focuses on the development of a new multimodal image registration algorithm for the study of stroke. Based on the T2-weighted magnetic resonance imaging (MRI) and diffusion-weighted imaging (DWI) of the rat brain, the proposed algorithm registers the infarct segmentation images to the atlas of the rat brain, where the percentage of the infarct areas in some specific brain regions can be calculated. The concepts of the B-spline and fluid registration methods are combined to our non-rigid image registration algorithm, which adopts a stepwise sampling technique to improve the registration efficiency. The proposed method was evaluated using various simulated and clinical images. By comparing with other competing methods, the advantages of our method were demonstrated based on the registered images and their evaluation scores. It was suggested that the proposed image registration scheme is potential to facilitate subsequent stroke-related studies.