EpiEEG-60: A Balanced Epileptic Seizure and Normal EEG Dataset

This dataset contains raw Electroencephalogram (EEG) recordings collected from 60 distinct individuals. The cohort is equally divided into two groups: 30 neurologically normal subjects and 30 abnormal subjects clinically diagnosed with epilepsy. The dataset is intended to support research in automated seizure detection, neurological signal processing, and machine learning classification tasks.

#Participants and Demographics#

Total Subjects: 60
Groups: 30 Normal, 30 Epilepsy
Demographics: The participants include both male and female subjects with an age range of 17 to 70 years.
Validation: All recordings were manually labeled and verified by an expert neurologist from the National Institute of Neuroscience and Hospital (NINS).

#Data Acquisition Details#

Instrumentation: Signals were acquired using a Nihon Kohden EEG system.
Sampling Rate: 500 Hz.
Recording Duration: Approximately 30 ± 2 minutes per subject.
Channel Configuration: Each recording comprises 43 channels, including standard EEG (electroencephalogram), EOG (electro-oculogram), MISC (miscellaneous reference), and BIO (physiological) channels.

Filtering: Default acquisition settings included a 0–250 Hz bandwidth filter.

#File Formats and Structure#
The dataset is organized using a structured directory hierarchy. The main directory, Ready_data, contains two subdirectories: Normal and Abnormal. Inside these, subject-specific folders (e.g., Patient_01, Patient_02) contain the recording sessions.
Each recording session includes seven proprietary Nihon Kohden files necessary for full decoding and interpretation:

.EEG: Contains the continuous, digitized raw EEG voltage waveforms.
.21E: Header file providing essential acquisition parameters (channel indexing, sensitivity, sampling frequency) required for parsing the .EEG file.
.EVT: Event annotation file storing time-stamped markers for clinical events (e.g., "Eyes Open," "Hyperventilation," "Seizure Onset").
.CMT: Electrode configuration file defining channel labels and 10–20 system positions.
.CN2: Channel-specific settings (amplifier configuration, calibration factors).
.LOG: System log recording session history and hardware status.
.TRD: Trend data files containing compressed signal views.

#Usage Notes#
These files are binary and typically require specialized libraries for processing. Python users can utilize MNE-Python (using the read_raw_nihon function) to load the data, as the .21E and .EEG files function together to reconstruct the signal. The .EVT files are critical for extracting epochs related to specific stimuli or clinical states.

本数据集包含来自60名不同个体的原始脑电图（Electroencephalogram, EEG）记录。该队列被平均分为两组：30名神经系统健康受试者与30名经临床诊断为癫痫的异常受试者。本数据集旨在支持自动化癫痫发作检测、神经信号处理以及机器学习分类任务相关研究。

#受试者与人口统计学特征#

受试者总数：60名
分组：30名健康组，30名癫痫组
人口统计学特征：受试者涵盖男性与女性，年龄跨度为17至70岁。
标注验证：所有记录均由神经科学与国家医院（National Institute of Neuroscience and Hospital, NINS）的专业神经内科医师进行人工标注与审核。

#数据采集细节#

采集设备：信号采集采用Nihon Kohden脑电图采集系统。
采样率：500 Hz。
记录时长：每名受试者的记录时长约为30±2分钟。
通道配置：每份记录包含43个通道，涵盖标准脑电图、眼电图（electro-oculogram, EOG）、杂项参考通道（miscellaneous reference, MISC）以及生理通道（physiological, BIO）。

滤波设置：默认采集参数包含0–250 Hz带宽的滤波。

#文件格式与组织结构#
本数据集采用结构化目录层级进行组织。主目录为Ready_data，包含两个子目录：Normal与Abnormal。在这两个子目录下，以受试者命名的文件夹（例如Patient_01、Patient_02）中存储对应受试者的记录会话。
每个记录会话包含7个Nihon Kohden专属格式的文件，这些文件是完整解码与解析记录所需的全部文件：

.EEG：存储连续的数字化原始脑电电压波形。
.21E：头文件，提供解析.EEG文件所需的关键采集参数，包括通道索引、灵敏度与采样频率。
.EVT：事件标注文件，存储带有时间戳的临床事件标记（例如"Eyes Open"、"Hyperventilation"、"Seizure Onset"）。
.CMT：电极配置文件，用于定义通道标签与10-20系统电极位置。
.CN2：通道专属配置文件，包含放大器配置与校准系数。
.LOG：系统日志文件，记录会话历史与硬件运行状态。
.TRD：趋势数据文件，存储压缩后的信号概览。

#使用说明#
本数据集文件均为二进制格式，通常需要使用专用库进行处理。Python用户可借助MNE-Python工具（通过read_raw_nihon函数）加载数据，因.21E与.EEG文件可协同完成信号重构。.EVT文件对于提取与特定刺激或临床状态相关的脑电片段至关重要。