[论文解读] Evaluation of distortion correction methods in diffusion MRI of the spinal cord
本研究在3T扫描仪上使用反转梯度极性的方法,评估了四种畸变校正技术——HySCO、Block-Matching、TOPUP和Voss在脊髓扩散MRI中的表现。通过几何度量指标,如扩散方向的方位集中度和与解剖图像的互相关性,HySCO展现出更优的校正性能,显著提升了图像与脊髓中心线的对齐程度及图像保真度,尤其在扫描窗口的头侧和尾侧区域表现更佳。
Background: Magnetic field inhomogeneities generate important geometric distortions in reconstructed echo-planar images. Various procedures were proposed for correcting these distortions on brain images; yet, few neuroimaging studies tailored and incorporated the use of these techniques in spinal cord diffusion MRI. Purpose: We present a comparative evaluation of distortion correction methods that use the reversed gradient polarity technique on spinal cord. We propose novel geometric metrics to measure the alignment of the reconstructed diffusion model with the apparent centerline of the spinal cord. Subjects: 95 subjects, among which 29 healthy controls and 66 multiple sclerosis patients. Assessment: Geometric distortions were corrected using 4 state-of-the-art methods. We measured the alignment of the principal direction of diffusion with the apparent centerline of the spine after correction and the correlation with the reference anatomical image. Results are computed per vertebral level, to evaluate the impact on different portions of the spine. Besides, subjective evaluation of the quality of the correction of healthy subjects images was performed by three expert raters. Results: As a result of distortion correction, the diffusion directions are better aligned locally with the centerline, in particular at both ends of the acquisition window. The cross-correlation with anatomical image is also improved by Hyperelastic Susceptibility Artefact Correction (HySCO) and block-matching. The subjective evaluation for HySCO is significantly better (p < 0.05) than for Block-Matching; TOPUP performs significantly worse than the three other methods. Conclusion: Correction based on HySCO provide best results among the selected methods.
研究动机与目标
- 评估并比较基于反转梯度极性的最新脊髓扩散MRI畸变校正方法。
- 解决由B0不均匀性引起的几何畸变,此类畸变在脊髓中尤为严重,因其体积小且周围组织复杂。
- 提出并应用新型几何度量指标,量化扩散方向与脊髓表观中心线之间的对齐程度。
- 通过专家评分员进行主观评估,以评价校正后图像的视觉质量。
- 确定哪种方法在不同脊椎节段中均能提供最准确、最可靠的畸变校正效果。
提出的方法
- 从95名受试者(29名健康对照,66名多发性硬化患者)获取3T扩散加权回波平面成像(EPI)数据,b值为900 s/mm²,共30个扩散方向。
- 应用四种畸变校正方法:超弹性磁 susceptibility 艺术品校正(HySCO)、Block-Matching、TOPUP和Voss,所有方法均基于反转梯度极性。
- 提出一种新型几何度量指标——扩散方向的方位集中度(ACD),用于衡量其与脊髓表观中心线的对齐程度。
- 使用校正后的扩散数据与T2加权解剖图像之间的互相关性和互信息作为额外的保真度度量指标。
- 对脊髓体积进行体素级分割,并按每个脊椎节段(C1–T2)计算度量指标,以评估校正性能在不同区域的差异。
- 通过基于网络的界面开展主观评估,三位专家评分员对四种方法的图像质量进行排序,采用包含评分员随机效应的逻辑回归模型进行分析。
实验结果
研究问题
- RQ1在不同脊椎节段中,哪种畸变校正方法最有效地将主扩散方向与脊髓表观中心线对齐?
- RQ2不同校正方法如何影响图像保真度?保真度通过与T2加权解剖图像的互相关性和互信息进行衡量。
- RQ3畸变校正的性能在脊髓的不同区域(如颈椎、胸椎或脑干)是否存在显著差异?
- RQ4专家评分员对图像质量的主观排名与客观几何度量及基于强度的度量结果相比如何?
- RQ5HySCO、Block-Matching、TOPUP和Voss方法之间的校正性能是否存在统计学上的显著差异?
主要发现
- 与未校正数据相比,HySCO显著提高了扩散方向的方位集中度(ACD)(p < 10⁻³),且在脑干和C1节段表现尤为突出。
- HySCO在与T2加权解剖图像的互相关性方面达到最高(均值0.242),与未校正数据相比p值为2.1×10⁻⁴,表明其图像对齐效果更优。
- 主观评估显示,HySCO显著优于Block-Matching(p < 0.05),而TOPUP在所有方法中表现最差。
- Block-Matching和HySCO在互相关性方面均有显著提升(p值分别为0.029和2.1×10⁻⁴),而TOPUP和Voss未达到显著性水平。
- 互信息分析进一步证实HySCO的优越性(p = 4.1×10⁻¹⁴),同时显示Block-Matching表现强劲(p = 10⁻⁵),表明其结构相似性更优。
- 不同校正方法之间观察到显著的脊髓体积差异(p < 0.05),其中HySCO和Block-Matching在分割中表现出更高的稳定性和准确性。
更好的研究,从现在开始
从阅读论文到最终审阅,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。