Aglieri Virginia, Cagna Bastien, Belin Pascal, Takerkart Sylvain
Institut des Neurosciences de la Timone, UMR 7289, CNRS and Université Aix-Marseille, 13005 Marseille, France.
International Laboratories for Brain, Music and Sound Research, Montreal, QC H2V 2S9, Canada.
Data Brief. 2020 Jan 25;29:105170. doi: 10.1016/j.dib.2020.105170. eCollection 2020 Apr.
Multivariate pattern analysis (MVPA) of functional neuroimaging data has emerged as a key tool for studying the cognitive architecture of the human brain. At the group level, we have recently demonstrated the advantages of an under-exploited scheme that consists in training a machine learning model on data from a set of subjects and evaluating its generalization ability on data from unseen subjects (see [1]). We here provide a data set that is fully ready to perform inter-subject pattern analysis, which includes 5616 single-trial brain activation maps recorded in 39 participants who were scanned using functional magnetic resonance imaging (fMRI) with a voice localizer paradigm. This data set should therefore reveal valuable for data scientists developing brain decoding algorithms as well as cognitive neuroscientists interested in voice perception.
功能神经影像数据的多变量模式分析（MVPA）已成为研究人类大脑认知结构的关键工具。在群体层面，我们最近展示了一种未充分利用的方案的优势，该方案包括在一组受试者的数据上训练机器学习模型，并在来自未见过的受试者的数据上评估其泛化能力（见[1]）。我们在此提供一个完全准备好进行受试者间模式分析的数据集，其中包括39名参与者记录的5616个单次试验脑激活图，这些参与者使用功能磁共振成像（fMRI）和语音定位范式进行扫描。因此，该数据集对于开发脑解码算法的数据科学家以及对语音感知感兴趣的认知神经科学家来说应该具有重要价值。
