无需配对数据，用自监督方法实现多模态体积成像的各向同性重建
CRIS: Cross-Plane Self-Supervised Isotropic Restoration for Anisotropic Volumetric Imaging Across Modalities
- 通过正交重构成像的2D条纹补全任务，实现跨平面自监督训练
- 在8倍各向异性条件下，脑MRI重建峰值信噪比达32.921 dB，分割一致性最优
- 适用于临床MRI与电镜数据，无需重新训练，可应对多种采样间隙
各向异性体积采集在临床MRI和体电子显微镜（vEM）中常见，其稀疏的层间采样导致切片变厚，影响正交重构成像与下游分析。本文提出CRIS，一种无需成对各向同性真值的跨平面自监督恢复框架。将3D恢复问题转化为对各向同性网格正交重构成像的2D条纹补全：高分辨率横截面被合成降质并周期性掩码用于训练；推理时空白切片构成各向同性网格，两个正交重构成像被恢复，再通过多视角平均融合。在两个MRI队列和两个显微镜基准上评估，最高支持8倍各向异性。在脑MRI上，CRIS达到32.921 ± 0.436 dB PSNR 和 0.963 ± 0.003 SSIM，优于插值、ECLARE、SMORE4、SIMPLE、SA-INR 和 ATME；分割一致性最佳（Dice 0.940 ± 0.004，ASSD 0.245 ± 0.014 mm，HD99 1.275 ± 0.061 mm）。在无参考腹部MRI上，FID/KID降至48.71/0.023，优于多种基线方法。在vEM数据上，于EPFL数据集上4倍和8倍各向异性下分别达到29.100 dB/0.830和26.874 dB/0.722的3D PSNR/SSIM；在噪声血脑数据上为21.935 ± 0.437 dB / 0.696 ± 0.024。鲁棒性实验表明，单一模型在3–7倍间隙及不同方向降质下仍优于插值（PSNR 36.36–31.14 dB vs. 33.07–27.85 dB，SSIM 0.977–0.932 vs. 0.951–0.853）。结果表明，CRIS是一种无需配对目标或特定配置重训练的通用各向同性恢复路径。代码已公开于https://github.com/adi-hatav/CRIS。
原文摘要 · Abstract (English)
Anisotropic volumetric acquisitions are common in clinical MRI and volume electron microscopy (vEM), where sparse through-plane sampling creates thick slices or sections that degrade orthogonal reformats and downstream analysis. We present CRIS, a cross-plane self-supervised framework for isotropic restoration without paired isotropic ground truth. CRIS casts 3D restoration as 2D stripe completion on orthogonal reformats of an isotropic grid: high-resolution in-plane slices are synthetically degraded and periodically masked for training, while at inference blank slices define the isotropic grid, two orthogonal reformats are restored, and predictions are fused by multi-view averaging. We evaluate CRIS on two MRI cohorts and two microscopy benchmarks up to 8x anisotropy. On brain MRI, CRIS achieves 32.921 +/- 0.436 dB PSNR and 0.963 +/- 0.003 SSIM, outperforming interpolation, ECLARE, SMORE4, SIMPLE, SA-INR, and ATME, and gives the best segmentation consistency (Dice 0.940 +/- 0.004, ASSD 0.245 +/- 0.014 mm, HD99 1.275 +/- 0.061 mm). On reference-free abdominal MRI, CRIS reduces FID/KID to 48.71/0.023, outperforming interpolation, ECLARE, SMORE4, and SIMPLE. On vEM, CRIS achieves 29.100 dB/0.830 3D PSNR/SSIM at 4x and 26.874 dB/0.722 at 8x on EPFL, and 21.935 +/- 0.437 dB/0.696 +/- 0.024 on noisy hemibrain data. In a dedicated robustness experiment, one variable-gap CRIS model evaluated across gap factors 3-7 and coronal, axial, and sagittal degradations maintained higher PSNR/SSIM than interpolation (36.36-31.14 dB and 0.977-0.932 vs. 33.07-27.85 dB and 0.951-0.853). These results support CRIS as a modality-flexible route to isotropic restoration without paired isotropic targets or configuration-specific retraining. Code is available at https://github.com/adi-hatav/CRIS.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。