fMRI study of experimental endotoxemia in humans

This dataset was acquired at Hannover Medical School, Hanover, Germany. The study complied with the Declaration of Helsinki, and was approved by the local ethics committee (#7427). All subjects gave written informed consent, and consent to publish their data anonimously.

Using high-resolution functional magnetic resonance imaging (fMRI) we recorded brain activity in healthy male subjects undergoing experimental inflammation from intravenous endotoxin. Four fMRI runs covered key phases of the developing inflammation: pre-inflammatory baseline, onset of endotoxemia, onset of proinflammatory cytokinemia, and peak of proinflammatory cytokinemia. We informed the participants that they would either receive an endotoxin or saline injection at some point during the fMRI experiment. However, all subjects received the endotoxin with the start of the second fMRI acquisition.

General exclusion criteria included a body mass index of <18 and >30 kg/m², any concurrent medical condition, history of allergies, current use of prescription and non-prescription medications, smoking, and regular high alcohol use. To exclude any inflammatory diseases that may aggravate through the HEM, each subject was interviewed and examined by a physician before being admitted to the study.

Data acquisition
Experimental design [min]

-30 -> -10: Baseline fMRI

-5: Baseline blood levels

0: Endotoxin administration

0-20: Second fMRI

20, 30: Blood sampling

30-50: Third fMRI

50, 60, 70, 80: Blood sampling

80-100: Fourth fMRI

100, 110, 120, 180, 240, 300, 360: Blood sampling

Human endotoxemia model
All subjects received an intravenous bolus injection of endotoxin over one minute through an intravenous catheter in an antecubital forearm vein. GMP-grade lipopolysaccharide from Escherichia coli O:113:H10:K-strain (Lot 94332B1) provided by National Institute of Health Clinical Center, Bethesda, MD, USA was prepared for human use by reconstitution with sterile water for injection, shaking for 15 minutes on a vortex shaker and final dilution. We used a dose of 1ng/kg (0,02 ml/kg) body weight. All endotoxin solutions were administered immediately after their preparation. Subjects were injected between nine and ten o’clock in the morning and discharged six to eight hours later, when their symptoms had receded, all altered physiological parameters had demonstrated consistent reduction toward baseline values, and the physical exam was normal.

MRI data
All MR images were acquired on a Siemens 3T MAGNETOM Skyra using a 64-channel head/neck coil. The scanning protocol consisted of the following sequences (see Sequences.ods):

func_i: Functional whole brain gradient-echo echo-planar images (EPI) (TR=1180 ms; TE=32 ms; 2 mm isotropic resolution; simultaneous multi-slice factor=6; partial Fourier=7/8)

func_ref: Reference scan for motion correction and template formation; equivalent to func but without multi-band acceleration (TR=6770 ms)

SE_pe1_pe2: Reference scans for unwarping: Two spin-echo images matched to func in distortion without multi-band acceleration; one with the same, the other one with inverted phase encoding direction.

t1: T1-weighted magnetisation-prepared rapid acquisition gradient-echo image (MPRAGE) (TR=2400 ms; TE=2.13 ms; TI=1000 ms; 1 mm isotropic resolution; in-plane acceleration factor=2)

t2: T2-weighted image (TR=3200 ms; TE=564 ms; 1 mm isotropic resolution; in-plane acceleration factor=2)

fMRI data preprocessing
Our preprocessing pipleine (JPreprocessing) is optimised for the brainstem and hypothalamus by avoiding superfluous resampling steps and unnecessary smoothing. On that account, motion correction (MCFLIRT [Jenkinson et al., 2002]) and unwarping (topup [Andersson et al.,2003]) are applied in a single transformation. Afterwards, brain extraction (BET [Smith, 2002]), grand mean scaling and high pass filtering (0.005 Hz) are applied. The data are not smoothed.

Two study templates were generated using Advanced Normalization Tools (ANTs [Avants et al., 2008] using antsMultivariateTemplateConstruction2.sh). The first one using the unwarped EPI reference images (func_ref); the second one using the T1-images.

Physiological data
The following physiological measures were acquired with an MR-compatible BIOPAC MP150 system

Blood pressure (systolic and dyastolic): Non-invasive continuous blood pressure of the digital artery (pulse decomposition analysis using CareTaker)

Electrodermal activity

Photoplethysmography

Respiration (belt)

Electrocardiography

Subject information
01    m    25 a    175 cm    70 kg
02    m    44 a    192 cm    84 kg
03    m    27 a    188 cm    85 kg
04    m    19 a    183 cm    99 kg
05    m    20 a    183 cm    90 kg
06    m    26 a    180 cm    78 kg
07    m    21 a    189 cm    80 kg

Missing data

sub01: functional data during run3 and run4 is shorter

Data structure
data.csv

Contains blood parameters, symptom ratings, as well as blood pressure measurements for all subjects.

sequences.ods

MRI sequence parameters

subXX

acqparams.txt: Acquisition parameters needed for topup

func_i: raw functional data

func_ref: Reference image for motion correction and template generation

physio.acq: Physiological measurements

Trigger

Blood pressure (systolic and dyastolic)

Electrodermal activity

Photoplethysmogram

Respiration

Electrocardiogram

physio_cuts.txt: time points in the physio data that correspond to fMRI blocks (start-time end-time TR #volumes)

SE_pe1_pe2: auxiliary image for unwarping

slicetiming_i.txt: Slice timing information in seconds

t1: Defaced T1-weighted image (undefaced images were used for template generation)

t2: Defaced T1-weighted image

templates

EPI-template: Generated from unwarped func_ref images

Transformations for all subjects (can be applied using antsApplyTransforms)

wb_mask

EPI-2-T1: Transformation from EPI to T1-template

MNI-template: T1-template warped into MNI_152

wb_mask

T1-template: Generated from t1 images

hyp_mask

wb_mask

T1-2-MNI: Transformation from T1 to MNI_152-template

JPreprocessing
Preprocessing pipeline

本数据集采集自德国汉诺威的汉诺威医学院。本研究符合《赫尔辛基宣言》，并获得当地伦理委员会批准（编号：7427）。所有受试者均签署了书面知情同意书，并同意匿名发布其相关数据。

我们采用高分辨率功能磁共振成像（functional magnetic resonance imaging, fMRI）记录了接受静脉内毒素诱导实验性炎症的健康男性受试者的脑活动。四次fMRI扫描覆盖了炎症发展的关键阶段：炎症前基线期、内毒素血症发作期、促炎性细胞因子血症发作期以及促炎性细胞因子血症峰值期。我们告知受试者，在fMRI实验过程中的某个时间点，他们将接受内毒素或生理盐水注射。但实际上，所有受试者均在第二次fMRI扫描开始时接受了内毒素注射。

通用排除标准包括体重指数（BMI）<18或>30 kg/m²、任何并存内科疾病、过敏史、当前使用处方及非处方药物、吸烟以及规律大量饮酒。为排除可能因内毒素血症模型加重的炎症性疾病，所有受试者在入组前均接受了医师的问诊与体格检查。

数据采集
实验设计（单位：分钟）
-30 ~ -10：基线fMRI扫描
-5：基线血液采样
0：内毒素给药
0 ~ 20：第二次fMRI扫描
20、30：血液采样
30 ~ 50：第三次fMRI扫描
50、60、70、80：血液采样
80 ~ 100：第四次fMRI扫描
100、110、120、180、240、300、360：血液采样

人内毒素血症模型
所有受试者均通过前臂肘前静脉的静脉导管，在1分钟内静脉推注内毒素。所用试剂为美国马里兰州贝塞斯达市国立卫生研究院临床中心提供的大肠杆菌O:113:H10:K菌株的GMP级脂多糖（批号：94332B1），经注射用无菌水复溶、涡旋振荡15分钟后进行最终稀释，制备供人体使用的制剂。给药剂量为1ng/kg体重（0.02ml/kg体重），所有内毒素溶液均在制备完成后立即给药。受试者于上午9时至10时之间接受注射，并在症状消退、所有异常生理参数恢复至基线水平且体格检查正常后的6至8小时出院。

MRI数据
所有磁共振图像均采用西门子3T MAGNETOM Skyra扫描仪，搭配64通道头颈部线圈采集。扫描方案包含以下序列（详见Sequences.ods）：
func_i：全脑功能梯度回波-回波平面成像（Echo-planar Imaging, EPI）（TR=1180ms；TE=32ms；各向同性分辨率2mm；同时多层因子=6；部分傅里叶变换=7/8）
func_ref：用于运动校正与模板构建的参考扫描，参数与func_i一致，但未启用多波段加速（TR=6770ms）
SE_pe1_pe2：用于畸变校正的参考扫描：两张自旋回波图像，与func_i的畸变匹配且未启用多波段加速，分别采用正向与反向相位编码方向
t1：T1加权磁化准备快速梯度回波成像（MPRAGE）（TR=2400ms；TE=2.13ms；TI=1000ms；各向同性分辨率1mm；平面内加速因子=2）
t2：T2加权成像（TR=3200ms；TE=564ms；各向同性分辨率1mm；平面内加速因子=2）

fMRI数据预处理
我们的预处理流程（JPreprocessing）针对脑干与下丘脑进行了优化，通过避免不必要的重采样步骤与平滑操作实现。据此，运动校正（MCFLIRT [Jenkinson等, 2002]）与畸变校正（topup [Andersson等, 2003]）采用单次变换完成。随后进行脑提取（BET [Smith, 2002]）、总体均值缩放以及高通滤波（截止频率0.005Hz），未对数据进行平滑处理。
我们采用高级归一化工具（Advanced Normalization Tools, ANTs [Avants等, 2008]，使用antsMultivariateTemplateConstruction2.sh脚本）生成了两种研究模板：第一种基于去畸变的EPI参考图像（func_ref）；第二种基于T1加权图像。

生理数据
采用兼容MRI的BIOPAC MP150系统采集以下生理指标：
- 血压（收缩压与舒张压）：采用手指动脉无创连续血压监测（通过CareTaker进行脉搏分解分析）
- 皮肤电活动
- 光电容积描记法
- 呼吸（胸带式）
- 心电图

受试者信息
01 男 25岁 175cm 70kg
02 男 44岁 192cm 84kg
03 男 27岁 188cm 85kg
04 男 19岁 183cm 99kg
05 男 20岁 183cm 90kg
06 男 26岁 180cm 78kg
07 男 21岁 189cm 80kg

缺失数据
sub01：第三次与第四次扫描的功能数据时长较短

数据结构
data.csv
包含所有受试者的血液参数、症状评分以及血压测量数据。

sequences.ods
包含MRI序列参数

subXX
acqparams.txt：topup所需的采集参数
func_i：原始功能成像数据
func_ref：用于运动校正与模板生成的参考图像
physio.acq：生理测量数据
触发信号
血压（收缩压与舒张压）
皮肤电活动
光电容积描记图
呼吸信号
心电图
physio_cuts.txt：生理数据中与fMRI扫描块对应的时间点（格式：起始时间 结束时间 TR 扫描帧数）
SE_pe1_pe2：用于畸变校正的辅助图像
slicetiming_i.txt：以秒为单位的层时序信息
t1：去面部识别的T1加权图像（模板生成使用原始未去面部图像）
t2：去面部识别的T2加权图像

templates
EPI模板：基于去畸变的func_ref图像生成
所有受试者的变换矩阵（可通过antsApplyTransforms工具应用）
wb_mask
EPI到T1变换：从EPI模板到T1模板的空间变换
MNI模板： warped到MNI_152空间的T1模板
wb_mask
T1模板：基于t1图像生成
hyp_mask
wb_mask
T1到MNI变换：从T1模板到MNI_152模板的空间变换

JPreprocessing
预处理流程