关键词: 弥散张量成像, 图像配准, 逆向组合算法, 可变步长

Abstract: Diffusion Tensor Magnetic Resonance Imaging（DT-MRI） is a rising Magnetic Resonance Imaging（MRI） technology in recent years.As a non-invasive method to describe complex inner structure of human brain，it plays an important role in clinical diagnostics.Because the unfolding images from the tensor data of different people or the same people at different times are different，these images should be aligned into a unified framework.Inverse compositional algorithm is a power method for image registration.However，the inverse compositional algorithm has some flaws.The convergence rate will be slow when the error image between template image and target image is large.This paper proposes a variable step-size algorithm based on the inverse compositional algorithm，which can raise the convergence rate by changing the step size adaptively.Experimental results show that this improved algorithm is faster than the inverse compositional algorithm and robust to the geometry variation.

Key words: Diffusion Tensor Magnetic Resonance Imaging, image registration, inverse compositional algorithm, variable step-size