3D image registration is one of the most fundamental and computationally expensive operations in medical image analysis. Here, we present a mixed-precision, Gauss–Newton–Krylov solver for diffeomorphic registration of two images. Our work extends the publicly available CLAIRE library to GPU architectures. Despite the importance of image registration, only a few implementations of large deformation diffeomorphic registration packages support GPUs. Our contributions are new algorithms to significantly reduce the run time of the two main computational kernels in CLAIRE: calculation of derivatives and scattered-data interpolation. We deploy (i) highly-optimized, mixed-precision GPU-kernels for the evaluation of scattered-data interpolation, (ii) replace Fast-Fourier-Transform (FFT)-based first-order derivatives with optimized 8th-order finite differences, and (iii) compare with state-of-the-art CPU and GPU implementations. As a highlight, we demonstrate that we can register $256^3$ clinical images in less than 6 s on a single NVIDIA Tesla V100. This amounts to over 20$\times$ speed-up over the current version of CLAIRE and over 30$\times$ speed-up over existing GPU implementations.

3D 图像配准是医学图像分析中最基本且计算成本最高的操作之一。在这里，我们提出了一个混合精度的高斯-牛顿-克雷洛夫求解器，用于两个图像的微分同胚配准。我们的工作将公开可用的CLAIRE库扩展到 GPU 架构。尽管图像配准很重要，但只有少数大变形微分同胚配准包的实现支持 GPU。我们的贡献是新算法，可显着减少CLAIRE中两个主要计算内核的运行时间：导数计算和分散数据插值。我们部署 (i) 高度优化的混合精度 GPU 内核来评估分散数据插值，(ii) 用优化的 8 阶有限差分替换基于快速傅里叶变换 (FFT) 的一阶导数， (iii) 与最先进的 CPU 和 GPU 实现进行比较。作为一个亮点，我们证明我们可以注册$\mathrm{<span\; data-immersive-translate-walked="8eeee7e2-46b0-4118-9ec9-c3503e52f19b">25</span>}{\mathrm{<span\; data-immersive-translate-walked="8eeee7e2-46b0-4118-9ec9-c3503e52f19b">6</span>}}^{\mathrm{<span\; data-immersive-translate-walked="8eeee7e2-46b0-4118-9ec9-c3503e52f19b">3</span>}}$在单个 NVIDIA Tesla V100 上不到 6 秒即可生成临床图像。这总计超过20比当前版本的CLAIRE加速超过 30比现有 GPU 实现加速。