EEG-FM-Bench
收藏资源简介:
EEG-FM-Bench是一个全面的基准,用于系统地评估脑电图基础模型(EEGFMs)。该数据集由同济大学的研究人员创建,旨在解决脑电图分析中长期存在的挑战,如高个体间和个体内变异性、实验范式间的数据差异以及获取大规模、专家注释数据集的高成本。EEG-FM-Bench包括来自10个常见范式的14个公开数据集,涵盖了运动想象、睡眠分期、情绪识别、癫痫检测和阿尔茨海默病分类等领域。数据集大小、数据量、Tokens数等信息在论文中未提及。数据集创建过程包括任务和数据的精心策划、标准化数据处理流程和健壮的评价协议。该数据集的应用领域包括认知、情感和神经病理学等脑电图解码领域,旨在提高模型性能和泛化能力,促进科学进步。
EEG-FM-Bench is a comprehensive benchmark for systematically evaluating electroencephalography foundation models (EEGFMs). This dataset was developed by researchers from Tongji University to address long-standing challenges in EEG analysis, including high inter- and intra-individual variability, data discrepancies across experimental paradigms, and the high cost of acquiring large-scale expert-annotated datasets. EEG-FM-Bench includes 14 public datasets from 10 common paradigms, covering domains such as motor imagery, sleep staging, emotion recognition, epilepsy detection, and Alzheimer’s disease classification. Specific details such as dataset size, data volume, and number of Tokens are not mentioned in the paper. The dataset creation process involves meticulous curation of tasks and data, standardized data processing workflows, and a robust evaluation protocol. Its application fields span EEG decoding areas including cognition, emotion, and neuropathology, with the goal of improving model performance and generalization ability and promoting scientific advancement.




