Improved self-similarity-based multi-sequence magnetic resonance image registration method

Abstract

An improved multi-sequence magnetic resonance image registration method based on self-similarity, including the following steps: 1) for all sequences of three-dimensional magnetic resonance images, use non-local mean fast noise reduction algorithm to denoise layer by layer; 2) by The stacked Otsu algorithm extracts the foreground part, that is, the Otsu algorithm is used multiple times to superimpose the next foreground into the previous image; 3) Use local three-dimensional median filtering to optimize the result; 4) Use self-similarity Encode the preprocessed image; 5) Use a discrete-based optimization method to calculate the similarity measure between feature maps, and judge whether the similarity measure reaches the optimal state; 6) Use the final transformation matrix to transform the original floating image , to obtain the final result graph. The invention has strong anti-noise interference ability and high registration precision.

Description

technical field [0001] The invention relates to a multi-sequence magnetic resonance image registration method. Background technique [0002] Usually, the registration between different sequences of magnetic resonance images is to optimize the similarity measure through the similarity measure and optimization algorithm, and always find the best similarity measure value. The image registration method determines the degree of similarity between images of different sequences through the similarity measure, but the similarity measure ignores the structural information existing between the images. When the similarity measure between different sequences of magnetic resonance images falls into a local optimum , tends to output a different result than what a human visually should get. At present, there are many kinds of image registration, such as the research based on similarity measure, such as the image registration method based on mutual information, the mutual information image...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the shortcomings of the existing multi-sequence magnetic resonance image registration method in the case of noise, the conversion map effect of the multi-sequence magnetic resonance medical image is poor and the stability is poor, the invention provides a better stability. Noise strategy combined with a self-similarity-based transformation method to obtain multi-sequence MR transformation maps

Images