Brain magnetic resonance image registration method based on converter

Abstract

The invention provides a brain magnetic resonance image registration method based on a converter (TransBIR), and aims to solve the problem that the existing method based on a convolutional neural network (CNN) adopts a local parameter sharing mechanism with a limited receptive field to extract features, so that only local information in data can be extracted, and the global feature and position information are lost. In order to realize accurate registration of brain magnetic resonance images, a converter mechanism is introduced into a network to extract a long-distance dependency relationship between long-distance pixels, and a converter is used for replacing a traditional convolution filter to perform down-sampling, so that global features in data can be extracted to a great extent; and meanwhile, a position coding module is used for extracting position information, so that the network can obtain accurate global features, and the network is helped to realize accurate registration. According to the method, each region of interest can be well registered, so that accurate registration is realized.

Description

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine

Login to View More

AI Technical Summary

Benefits of technology

The technical effect of this patented technology is improved registration efficiency by utilizing an encoding/decision framework for converting images into digital form without limiting their ability to match accurately on different devices or environments. This results in higher-quality image processing capabilities compared to traditional methods like histograms and grayscale values.

Problems solved by technology

This patented technical problem addressed in this patents relates to improving the accuracy and efficiency of aligning nuclear medicine scans with patient healthcare records during neurosurgery procedures such as tumor resection or therapy monitoring. Current manually registered techniques require significant human effort and may result in variations across different individuals that affect their quality of life. Consequently there is a demand for faster and less invasively automated systems capable of accurately identifying these differences among individual subjects over longer periods of observation times while reducing error rates caused by factors like background noises.

Images