跳转至

文章背景与核心概要

神经影像学模态为表征大脑活动、结构和连接性提供了核心工具。当测量数据通过适当的估计或正则化建模为对称正定(SPD)值表示时,便会产生一个统一的视角。赋予了黎曼几何的SPD流形,为原则性的统计推断和机器学习提供了一个非欧几里得框架。

本文于2026年被《IEEE模式分析与机器智能汇刊》(TPAMI)接收,系统地综述了SPD矩阵学习框架内的分析与学习方法,将经典几何统计学与现代机器学习连接起来,广泛应用于神经影像和脑机接口(BCI)领域。


SPD矩阵学习用于神经影像分析:视角、方法与挑战 (SPD Matrix Learning for Neuroimaging Analysis: Perspectives, Methods, and Challenges)

概要 (Summary)

神经影像学模态为表征大脑活动、结构和连接性提供了必不可少的工具。当测量数据通过适当的估计或正则化建模为对称正定(SPD)值表示时,一个统一的视角便应运而生。配备了黎曼几何的SPD流形,为这些表示上的原则性统计推断和机器学习提供了一个非欧几里得框架。

本文于2026年被《IEEE模式分析与机器智能汇刊》(TPAMI)接收,系统地综述了SPD矩阵学习框架内的分析与学习方法,将神经影像与脑机接口(BCI)应用中的经典几何统计学与现代机器学习连接起来。

Neuroimaging modalities provide essential tools for characterizing brain activity, structure, and connectivity. When measurements are modeled as symmetric positive-definite (SPD)-valued representations through appropriate estimation or regularization, a unifying perspective emerges. The SPD manifold, endowed with Riemannian geometry, offers a non-Euclidean framework for principled statistical inference and machine learning.

This paper—accepted for publication in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) in 2026—systematically surveys analytical and learning approaches within an SPD matrix learning framework, connecting classical geometric statistics with modern machine learning across neuroimaging and brain-computer interface (BCI) applications.


论文元数据 (Paper Metadata)

元数据字段 (Metadata Field) 详情 (Details)
arXiv ID arXiv:2504.18882 [cs.LG]
相关 DOI 10.1109/TPAMI.2026.3726269
一级学科 机器学习 (cs.LG)
交叉学科 人工智能 (cs.AI)、图像与视频处理 (eess.IV)、神经元与认知 (q-bio.NC)
作者 Ce Ju, Reinmar Kobler, Antoine Collas, Motoaki Kawanabe, Cuntai Guan, Bertrand Thirion
发表状态 已被 IEEE TPAMI 接收 (2026)
提交历史 • v1: 2025年4月26日
• v2: 2026年1月7日
• v3: 2026年8月21日
Metadata Field Details
arXiv ID arXiv:2504.18882 [cs.LG]
Related DOI 10.1109/TPAMI.2026.3726269
Primary Subject Machine Learning (cs.LG)
Cross-Subjects Artificial Intelligence (cs.AI), Image and Video Processing (eess.IV), Neurons and Cognition (q-bio.NC)
Authors Ce Ju, Reinmar Kobler, Antoine Collas, Motoaki Kawanabe, Cuntai Guan, Bertrand Thirion
Publication Status Accepted in IEEE TPAMI (2026)
Submission History • v1: 26 Apr 2025
• v2: 7 Jan 2026
• v3: 21 Aug 2026

摘要 (Abstract)

神经影像学通过捕获大脑组织互补方面的模态,为表征大脑活动、结构和连接性提供了必不可少的工具。在这些多样化的模态中,当测量数据通过适当的估计或正则化程序建模为对称正定(SPD)值表示时,便会出现一个统一的视角。

由于具备黎曼几何,SPD流形为这些表示上的原则性统计推断和机器学习提供了一个非欧几里得框架。本综述在一个SPD矩阵学习框架内组织了这些分析和学习方法,将经典几何统计学与现代机器学习连接到神经影像和神经生理学应用中。我们系统地综述了从特定模态表示到几何浅层和深层学习范式的演进,突出了SPD矩阵学习如何在保持底层结构约束的同时,扩展到神经影像和脑机接口中的现代AI应用。

Neuroimaging provides essential tools for characterizing brain activity, structure, and connectivity through modalities that capture complementary aspects of brain organization. Across these diverse modalities, a unifying perspective arises when measurements are modeled as symmetric positive-definite (SPD)-valued representations through appropriate estimation or regularization procedures.

Endowed with Riemannian geometry, the SPD manifold provides a non-Euclidean framework for principled statistical inference and machine learning on these representations. This review organizes these analytical and learning approaches within a framework for SPD matrix learning that connects classical geometric statistics with modern machine learning across neuroimaging and neurophysiological applications. We systematically survey the progression from modality-specific representations to geometric shallow and deep learning paradigms, highlighting how SPD matrix learning preserves underlying structural constraints while extending to modern AI applications in neuroimaging and brain-computer interfaces.