Numerous filtering methods have been proposed for estimating asymmetric orientation distribution functions (ODFs) for diffusion magnetic resonance imaging (dMRI). It can be hard to make sense of all these different methods, which share similar features and result in similar outputs. In this work, we disentangle these many filtering methods proposed in the past and combine them into a novel, unified filtering equation. We also propose a self-supervised data-driven approach for calibrating the filtering parameter values. Our equation is implemented in an open-source GPU-accelerated python software to facilitate its integration into any existing dMRI processing pipeline. Our method is applied on multi-shell multi-tissue fiber ODFs from the Human Connectome Project dataset (1.25 ${\text{mm}}^3$ native resolution) and on single-shell single-tissue fiber ODFs from the Bilingualism and the Brain dataset (2.0 ${\text{mm}}^3$ isotropic resolution) to evaluate the occurence of asymmetric patterns on different spatial resolutions, representing cutting-edge and “clinical” research data. Asymmetry measures such as the asymmetric index (ASI) and our novel number of fiber directions (NuFiD) are then used to explain the behaviour of our method in these images. The contributions of this work are: (i) the disentanglement and unification of filtering methods for estimating asymmetric ODFs; (ii) a calibration method for automatically fixing the parameters governing the filtering; (iii) an open-source, efficient implementation of our unified filtering method for estimating asymmetric ODFs; (iv) a novel number of fiber directions (NuFiD) index for explaining asymmetric fiber configurations; and (v) a novel template of asymmetries, revealing that our filtering method estimates asymmetric configurations in at least 50% of the brain voxels ($\sim 31\text{\%}$ of the white matter and $\sim 63\text{\%}$ of the gray matter).

已经提出了多种滤波方法来估计扩散磁共振成像（dMRI）的不对称方向分布函数（ODF）。理解所有这些不同的方法可能很困难，因为它们具有相似的特征并产生相似的输出。在这项工作中，我们解开了过去提出的许多滤波方法，并将它们组合成一个新颖的、统一的滤波方程。我们还提出了一种自监督数据驱动方法来校准过滤参数值。我们的方程在开源 GPU 加速的 Python 软件中实现，以便于将其集成到任何现有的 dMRI 处理流程中。我们的方法应用于来自人类连接组项目数据集的多壳多组织纤维 ODF（1.25${\text{毫米}}^3$原始分辨率）以及来自双语和大脑数据集的单壳单组织纤维 ODF（2.0${\text{毫米}}^3$各向同性分辨率）来评估不同空间分辨率上不对称图案的出现，代表前沿和“临床”研究数据。然后使用不对称指数（ASI）和我们新颖的纤维方向数（NuFiD）等不对称测量来解释我们的方法在这些图像中的行为。这项工作的贡献是：（i）用于估计非对称 ODF 的滤波方法的解开和统一；(ii)自动固定控制过滤的参数的校准方法；(iii) 开源、高效地实施我们用于估计非对称 ODF 的统一过滤方法；(iv) 一种新颖的纤维方向数 (NuFiD) 指数，用于解释不对称纤维配置；(v) 一种新颖的不对称模板，揭示了我们的过滤方法估计了至少 50% 的大脑体素中的不对称配置（$～31\text{\%}$白质和$～63\text{\%}$灰质）。