A novel approach for Automatic Image Stitching of spinal cord MRI images using SIFT

This paper presents a novel approach for Automatic Image Stitching of spinal cord MRI Images by exploring the effectiveness of Scale Invariant Feature Transform(SIFT) for feature matching. Because of limitations of MRI machine, when a patient undergoes scan of spinal cord, three different images of cervical, thoracic and lumbar parts with some vertebral overlap are produced. However, it is desired to have single image of entire spinal cord for diagnosis. This paper aims at automatically generating single seamless image of spinal cord from three different images having some overlap. After initial set of feature correspondences has been computed from SIFT features, RANdom SAmple Consensus(RANSAC) is used to reject the outliers and robustly estimate the best fitting homography to produce high accuracy alignment despite of noisy correspondence between image pair. The experimental results are accessed objectively with the help of radiologists as well as doctors and are upto the mark from diagnostic perspective.