Unsupervised multi-modal image registration method based on integrated attention enhancement

Abstract

The invention discloses an unsupervised multi-modal image registration method based on integrated attention enhancement. The method comprises the following steps: 1, preprocessing a medical image; 2,designing a registration framework, constructing a convolutional neural network model, automatically learning network parameters by optimizing similarity measurement of image pairs, and directly estimating a deformation field of the image pairs; and 3, dividing the image data into a training set and a test set, training a network model by using the training set, and finally testing the test set byusing the trained network model. According to the method, the deformation parameters of the image pair are estimated by directly optimizing the target function through the deep learning technology, the method can adapt to different data for different data and has high generalization capacity, the cascade encoder and the EAM are designed to extract the characteristics useful for the registration task, and the registration performance is improved to a certain extent.

Description

technical field [0001] The invention relates to a registration method of medical images, in particular to an unsupervised multi-modal image registration method based on integrated attention enhancement, which belongs to the technical fields of medical image processing and deep learning. Background technique [0002] At present, image registration can be divided into traditional registration methods and deep learning methods. Traditional non-learning registration methods can be further divided into feature-based registration algorithms and grayscale-based registration algorithms. Firstly, the features of the reference image and the floating image are extracted, and then the corresponding relationship between the features is established through the matching strategy, and the deformation parameters of the image pair are obtained through feature matching. The result of this type of algorithm registration depends on the accuracy of feature extraction. If the extracted features ...

AI Technical Summary

Problems solved by technology

It overcomes the problems of iterative optimization and poor robustness in traditional registration methods. At the same time, it avoids the problem that supervised registration methods are difficult to obtain label data in practical applications by directly optimizing the objective function to estimate deformation parameters.

Images