Abstract
The purpose of medical image registration is to find geometric transformations that align two medical images so that the corresponding voxels on two images are spatially consistent. Nonrigid medical image registration is a key step in medical image processing, such as image comparison, data fusion, target recognition, and pathological change analysis. Existing registration methods only consider registration accuracy but largely neglect the uncertainty of registration results. In this work, a method based on the Bayesian fully convolutional neural network is proposed for nonrigid medical image registration. The proposed method can generate a geometric uncertainty map to calculate the uncertainty of registration results. This uncertainty can be interpreted as a confidence interval, which is essential for judging whether the source data are abnormal. Moreover, the proposed method introduces group normalization, which is conducive to the network convergence of the Bayesian neural network. Some representative learning-based image registration methods are compared with the proposed method on different image datasets. Experimental results show that the registration accuracy of the proposed method is better than that of the methods, and its antifolding performance is comparable to that of fast image registration and VoxelMorph. Furthermore, the proposed method can evaluate the uncertainty of registration results.
Highlights
Image registration is an image-processing process that aligns two or more images of the same scene captured at different times and different perspectives or by using different sensors [1, 2]
We optimized the parameters by the validation set and reported results in our test set. e predicted deformation field could not guarantee diffeomorphism; the transformation of irreversible regions caused an image to “fold” on itself
Advanced Normalization Tools (ANTs) (SyN) is a nonrigid registration method, and its registration accuracy was found to be higher than that of affine registration. e Dice scores of fast image registration (FAIM) slightly decreased as β increased, and its Dice scores were higher than those of VoxelMorph. e registration accuracy of Probab-Mul was slightly better than that of FAIM. e proposed method achieved the highest registration accuracy under all β values
Summary
Image registration is an image-processing process that aligns two or more images of the same scene captured at different times and different perspectives or by using different sensors [1, 2]. Nonrigid medical image registration is a key step in medical image processing In clinical diagnosis, it can judge a patient’s progress by aligning the brain magnetic resonance images of the patient with Alzheimer’s disease at different periods [3, 4]. Many optimization algorithms have been devised, such as gradient descent methods [21], conjugate gradient methods [22, 23], Powell’s conjugate direction method [24, 25], quasi-Newton methods [26, 27], Gauss–Newton method [28, 29], and stochastic gradient descent methods [30, 31] Similarity measurement methods, such as the sum of squared differences [32], the sum of absolute differences, cross-correlation [33], and mutual information [34], have been proposed
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
