Go Figure：神经影像透明度保留背景并澄清解释
神经元与认知
2025-04-11 v1
摘要
可视化对于传达科学结果至关重要。从历史上看，神经影像图仅描绘超过给定统计阈值的区域。这种做法严重偏倚了对结果和后续荟萃分析的解读，尤其是导致不可重复性。在此，我们提倡一种“透明阈值”方法，该方法不仅突出显示具有统计学显著性的区域，还包含提供关键实验背景的次阈值位置。这平衡了提炼建模结果和为现代神经影像学提供知情解读的双重需求。我们展示了四个示例，证明了透明阈值的诸多益处，包括：消除歧义、降低对非生理特征的过度敏感、捕捉潜在伪影、改善跨研究比较、减少不可重复性偏倚以及澄清解读。我们还演示了实现透明阈值的众多软件包，其中几个是作为这项工作的一部分最近添加或简化的。一场观点交锋讨论探讨了在与该领域研究人员的真实对话中提出的关于阈值的问题。我们希望通过展示透明阈值如何极大地改善神经影像学发现的解读（和可重复性），更多的研究人员将采用这种方法。
引用
@article{arxiv.2504.07824,
title = {Go Figure: Transparency in neuroscience images preserves context and clarifies interpretation},
author = {Paul A. Taylor and Himanshu Aggarwal and Peter Bandettini and Marco Barilari and Molly Bright and Cesar Caballero-Gaudes and Vince Calhoun and Mallar Chakravarty and Gabriel Devenyi and Jennifer Evans and Eduardo Garza-Villarreal and Jalil Rasgado-Toledo and Remi Gau and Daniel Glen and Rainer Goebel and Javier Gonzalez-Castillo and Omer Faruk Gulban and Yaroslav Halchenko and Daniel Handwerker and Taylor Hanayik and Peter Lauren and David Leopold and Jason Lerch and Christian Mathys and Paul McCarthy and Anke McLeod and Amanda Mejia and Stefano Moia and Thomas Nichols and Cyril Pernet and Luiz Pessoa and Bettina Pfleiderer and Justin Rajendra and Laura Reyes and Richard Reynolds and Vinai Roopchansingh and Chris Rorden and Brian Russ and Benedikt Sundermann and Bertrand Thirion and Salvatore Torrisi and Gang Chen},
journal= {arXiv preprint arXiv:2504.07824},
year = {2025}
}
备注
Approx 27 pages for main text, with 7 figures, and additional Supplementary section included