School of Automation, Northwestern Polytechnical University, Xi'an, China.
Cortical Architecture Imaging and Discovery Lab, Department of Computer Science and Bioimaging Research Center, The University of Georgia, Athens, GA, Georgia.
Brain Imaging Behav. 2018 Jun;12(3):743-757. doi: 10.1007/s11682-017-9733-8.
Recently, a growing body of studies have demonstrated the simultaneous existence of diverse brain activities, e.g., task-evoked dominant response activities, delayed response activities and intrinsic brain activities, under specific task conditions. However, current dominant task-based functional magnetic resonance imaging (tfMRI) analysis approach, i.e., the general linear model (GLM), might have difficulty in discovering those diverse and concurrent brain responses sufficiently. This subtraction-based model-driven approach focuses on the brain activities evoked directly from the task paradigm, thus likely overlooks other possible concurrent brain activities evoked during the information processing. To deal with this problem, in this paper, we propose a novel hybrid framework, called extendable supervised dictionary learning (E-SDL), to explore diverse and concurrent brain activities under task conditions. A critical difference between E-SDL framework and previous methods is that we systematically extend the basic task paradigm regressor into meaningful regressor groups to account for possible regressor variation during the information processing procedure in the brain. Applications of the proposed framework on five independent and publicly available tfMRI datasets from human connectome project (HCP) simultaneously revealed more meaningful group-wise consistent task-evoked networks and common intrinsic connectivity networks (ICNs). These results demonstrate the advantage of the proposed framework in identifying the diversity of concurrent brain activities in tfMRI datasets.
最近，越来越多的研究表明，在特定任务条件下，大脑同时存在多种活动，例如任务诱发的主导反应活动、延迟反应活动和内在大脑活动。然而，目前主流的基于任务的功能磁共振成像（tfMRI）分析方法，即广义线性模型（GLM），可能难以充分发现这些多样化和并发的大脑反应。这种基于减法的模型驱动方法侧重于直接从任务范式中引发的大脑活动，因此可能忽略了在信息处理过程中可能引发的其他可能的并发大脑活动。为了解决这个问题，在本文中，我们提出了一种新的混合框架，称为可扩展监督字典学习（E-SDL），以探索任务条件下的多样化和并发大脑活动。E-SDL 框架与以前的方法的一个关键区别在于，我们系统地将基本任务范式回归器扩展到有意义的回归器组中，以解释大脑信息处理过程中可能的回归器变化。在五个来自人类连接组计划（HCP）的独立且公开可用的 tfMRI 数据集上的应用同时揭示了更有意义的、群组一致的任务诱发网络和常见的内在连通性网络（ICN）。这些结果表明，该框架在识别 tfMRI 数据集中并发大脑活动的多样性方面具有优势。
