Efficient Deep Learning Models for Predicting Individualized Task Activation From Resting-State Functional Connectivity

Soren J. Madsen

doi:10.1002/hbm.70557

2026-06-18

DOI：10.1002/hbm.70557

原文链接

查看原文

摘要

En 中文

深度学习模型已展现出从静息态功能磁共振成像预测任务诱发的脑激活的潜力，为无需任务数据的个性化脑图谱绘制提供了途径。本研究系统评估了提升此类模型效率与可扩展性的架构策略。使用人类连接组计划的数据，我们复现了BrainSurfCNN框架，并引入两种扩展：BrainSERF，通过squeeze-and-excitation模块融入通道注意力机制；以及BrainSurfGCN，一种基于图模型，利用皮层网格拓扑进行高效信息传递。在空间相关性、Dice评分、Dice AUC和被试识别准确率等多重评估指标下，所有模型均达到可比的预测性能。尽管准确率相似，所提模型各有优势：BrainSERF在捕捉个体特异性特征上略有提升，而BrainSurfGCN在模型尺寸和训练时间上实现显著缩减，凸显了性能与计算效率间的有利权衡。除架构比较外，我们还探究了驱动预测准确率变异的因素。研究发现，行为任务表现、静息态数据质量以及任务激活的个体间变异性共同限制了预测保真度。特别是信号可靠性较低、变异性较高的对比任务在所有模型中均表现出可预测性降低。综合来看，这些发现表明整合拓扑和功能结构先验可提升深度学习模型的效率而无需牺牲准确率，同时强调预测性能本质上受限于基础神经信号的可靠性。

Keyword:

deep learning

functional MRI

resting-state

task contrast

AI总结

对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。

期刊

Human Brain Mapping

IF:

3.3

论文数:

6.8K

被引数:

2.6W

机构

Washington University in St. Louis

学者数:

1.0K

论文数: 425

被引数: 0

weill cornell medicine

学者数:

2.2K

论文数: 1.0K

被引数: 0

stanford university

学者数:

1.1W

论文数: 4.3K

被引数: 0

University of California

学者数:

9.7K

论文数: 3.6K

被引数: 8.3W

university of minnesota

学者数:

3.8K

论文数: 1.7K

被引数: 0

查看更多机构

引用论文

Task-induced brain connectivity promotes the detection of individual differences in brain-behavior relationships

NeuroImage

IF0

2020-02-01

OAAI

Rongtao Jiang; Nianming Zuo; Judith M. Ford; Shile Qi; Dongmei Zhi; Chuanjun Zhuo; Yong Xu; Zening Fu; Juan Bustillo; Jessica A. Turner; Vince D. Calhoun; Jing Sui

Visualising inter-subject variability in fMRI using threshold-weighted overlap maps

SCIENTIFIC REPORTS

IF3.9

2016-02-05

OAAI

Seghier, Mohamed L.; Price, Cathy J.

The NIMH Research Domain Criteria (RDoC) Project: Precision Medicine for Psychiatry

AMERICAN JOURNAL OF PSYCHIATRY

IF14.7

2014-04-01

1.1K

PREAI

Insel, Thomas R.

The minimal preprocessing pipelines for the Human Connectome Project 人类连接体项目的最小预处理管道

NeuroImage

IF0

2013-10-01

OAAI

Matthew F. Glasser; Stamatios N. Sotiropoulos; J. Anthony Wilson; Timothy S. Coalson; Bruce Fischl; Jesper L. Andersson; Junqian Xu; Saad Jbabdi; Matthew Webster; Jonathan R. Polimeni; David C. Van Essen; Mark Jenkinson

Dissecting task-based fMRI activity using normative modelling: an application to the Emotional Face Matching Task

COMMUNICATIONS BIOLOGY

IF5.1

2024-07-20

OAAI

Savage, Hannah S.; Mulders, Peter C. R.; van Eijndhoven, Philip F. P.; van Oort, Jasper; Tendolkar, Indira; Vrijsen, Janna N.; Beckmann, Christian F.; Marquand, Andre F.

Structural Connectivity Fingerprints Predict Cortical Selectivity for Multiple Visual Categories across Cortex

Cerebral Cortex

IF0

2015-01-26

OAAI

David E. Osher; Rebecca R. Saxe; Kami Koldewyn; John D.E. Gabrieli; Nancy Kanwisher; Zeynep M. Saygin

Squeeze-and-Excitation Networks 挤压和激励网络

2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition

IF0

2018-06-01

OAAI

Jie Hu; Li Shen; Gang Sun

Predicting an individual’s dorsal attention network activity from functional connectivity fingerprints

Journal of Neurophysiology

IF0

2019-07-01

OAAI

David E. Osher; James A. Brissenden; David C. Somers

Task-free MRI predicts individual differences in brain activity during task performance

Science

IF0

2016-04-08

OAAI

I. Tavor; O. Parker Jones; R. B. Mars; S. M. Smith; T. E. Behrens; S. Jbabdi

Function in the human connectome: Task-fMRI and individual differences in behavior 人类连接体中的功能: 任务功能磁共振成像和行为的个体差异

NEUROIMAGE

IF4.5

2013-10-01

1.1K

OAAI

Barch, Deanna M.; Burgess, Gregory C.; Harms, Michael P.; Petersen, Steven E.; Schlaggar, Bradley L.; Corbetta, Maurizio; Glasser, Matthew F.; Curtiss, Sandra; Dixit, Sachin; Feldt, Cindy; Nolan, Dan; Bryant, Edward; Hartley, Tucker; Footer, Owen; Bjork, James M.; Poldrack, Russ; Smith, Steve; Johansen-Berg, Heidi; Snyder, Abraham Z.; Van Essen, David C.

查看更多内容