Notes: Principles of fMRI 1 (Coursera)
2015-10-18 14:26
295 查看
course link: https://class.coursera.org/fmri1-001
Part 1
❤ Three fundmental goals in fMRI:
localization (brain mapping approach: task comparison, brain-behavior correlation, information-based mapping);
connectivity (functional connectivity (seed-based), effectivity connectivity (DCM), multivariate connectivity (ICA, PCA, graph theory));
prediction (use the brain activities to predict something such as behavior)
❤ fMRI data structure:
TR: temporal resolution
Sturctural images:T1, WM<GM<CSF (longitudinal relaxation)
Functional images: T2* images
T2: teanscerse reliaxation
❤ fmri data structure
Field of view (FOV): to what extent of view in each direction we can see the brain
Slice thickness
eg. If the FOC is 192 mm, matrix size is 64 mm (the area of each slice), the slice thickness is 3 mm, then voxel size: 192mm (FOV)/64 (matrix size)*3 (slice thickness) = 3*3*3 mm voxel size (the last 3 means slice thickness)
hierarchy: Experiment-subjects-session-run-volume-slice-voxel
❤ Statistical map: the colors indicate reliable, non-zero effects
❤ Reverse inference: observed brain actives -> the feeling
One fallacy: if P->Q, Q, therefore P (true only if P is the only factor leads to Q)
❤ For regional brain activation to have high positive predictive value:
- It must respond consistently to the task/state (high sentivitivity);
- It must respond only to the task/state (high specificity).
Part 2
course link: https://class.coursera.org/fmri1-001
❤ T1 time
WM = 600;
GM = 1000;
CSF = 3000.
❤ Terms relating to time
TR: how often we excite the nuclei;
TE: how soon after exciation we begin data collection.
❤ K-space
K-space is frequency space.
Doing inverse Fourier transformation on points in k-space could reconstruct brain image we need.
Each individual point in image space depends on all points in k-space (in other words, the value of the points in k-space tells us its relative contribution in reconstruct the brain image).
"Low spatial frequencies represent parts of the object that change in a spatially slow manner (Contrast); high spatial frequencies represent small structures whose size is on the same order as the voxel size (Tissue boundaries)." -- the center points in the k-space contribute most of the brain image, while the outskirt part mostly contribute to the sketch of the brain.
Part 3
❤ BOLD signal
- Oxyhemoglobin is diamagnetic; while deoxyhemoglobin is paramagnetic (distort the magetic fields, and suppress the MR signal - when deoxyhemoglobin decrease, then the T2*-weighted signal increase).
- Bold signal often corresponds relatively closely to the local field potential (LFP - often reflect the integrated post-synaptic activiy across a group of neurons) - the electrical field potentional surrounding a group of cells.
- Bold signal does not always reflect changes in neuron activity.
❤ Basic quality control
- SNR (signal-to-noise ratio): a basic measure of effect size.
- CNR (contrast-to-noise ratio)
The two measurements can be calculated at both spatial and temporal level. Note that for temporally detrended data, the temporal SNR (or functional SNR) is also called Signal-to-Fluctuation-Noise Ratio (SFNR). Temporal CNR is also signal sentivity.
- Bold response is non-linear, which may cause nonlinear 'saturation'. This effect can be reduced by increasing the intervals between two stimuli.
❤ fMRI artefacts and non-signal-related noise
- Drift (low frequency noise)
Slow changes in voxel intersity over time - scanner instabilities and aliased physiological noise. Should be taken care about when pre-process images and conduct statistical analysis.
To aviod drift the experiment design should be fast.
-Motion
Should be taken care about when pre-process images and conduct statistical analysis. Only motion correction in pre-processing steps is not enough.
- Respiration and heart rate
TR that is too low may give rise to aliasing.
Part 4
Part 1
❤ Three fundmental goals in fMRI:
localization (brain mapping approach: task comparison, brain-behavior correlation, information-based mapping);
connectivity (functional connectivity (seed-based), effectivity connectivity (DCM), multivariate connectivity (ICA, PCA, graph theory));
prediction (use the brain activities to predict something such as behavior)
❤ fMRI data structure:
TR: temporal resolution
Sturctural images:T1, WM<GM<CSF (longitudinal relaxation)
Functional images: T2* images
T2: teanscerse reliaxation
❤ fmri data structure
Field of view (FOV): to what extent of view in each direction we can see the brain
Slice thickness
eg. If the FOC is 192 mm, matrix size is 64 mm (the area of each slice), the slice thickness is 3 mm, then voxel size: 192mm (FOV)/64 (matrix size)*3 (slice thickness) = 3*3*3 mm voxel size (the last 3 means slice thickness)
hierarchy: Experiment-subjects-session-run-volume-slice-voxel
❤ Statistical map: the colors indicate reliable, non-zero effects
❤ Reverse inference: observed brain actives -> the feeling
One fallacy: if P->Q, Q, therefore P (true only if P is the only factor leads to Q)
❤ For regional brain activation to have high positive predictive value:
- It must respond consistently to the task/state (high sentivitivity);
- It must respond only to the task/state (high specificity).
Part 2
course link: https://class.coursera.org/fmri1-001
❤ T1 time
WM = 600;
GM = 1000;
CSF = 3000.
❤ Terms relating to time
TR: how often we excite the nuclei;
TE: how soon after exciation we begin data collection.
❤ K-space
K-space is frequency space.
Doing inverse Fourier transformation on points in k-space could reconstruct brain image we need.
Each individual point in image space depends on all points in k-space (in other words, the value of the points in k-space tells us its relative contribution in reconstruct the brain image).
"Low spatial frequencies represent parts of the object that change in a spatially slow manner (Contrast); high spatial frequencies represent small structures whose size is on the same order as the voxel size (Tissue boundaries)." -- the center points in the k-space contribute most of the brain image, while the outskirt part mostly contribute to the sketch of the brain.
Part 3
❤ BOLD signal
- Oxyhemoglobin is diamagnetic; while deoxyhemoglobin is paramagnetic (distort the magetic fields, and suppress the MR signal - when deoxyhemoglobin decrease, then the T2*-weighted signal increase).
- Bold signal often corresponds relatively closely to the local field potential (LFP - often reflect the integrated post-synaptic activiy across a group of neurons) - the electrical field potentional surrounding a group of cells.
- Bold signal does not always reflect changes in neuron activity.
❤ Basic quality control
- SNR (signal-to-noise ratio): a basic measure of effect size.
- CNR (contrast-to-noise ratio)
The two measurements can be calculated at both spatial and temporal level. Note that for temporally detrended data, the temporal SNR (or functional SNR) is also called Signal-to-Fluctuation-Noise Ratio (SFNR). Temporal CNR is also signal sentivity.
- Bold response is non-linear, which may cause nonlinear 'saturation'. This effect can be reduced by increasing the intervals between two stimuli.
❤ fMRI artefacts and non-signal-related noise
- Drift (low frequency noise)
Slow changes in voxel intersity over time - scanner instabilities and aliased physiological noise. Should be taken care about when pre-process images and conduct statistical analysis.
To aviod drift the experiment design should be fast.
-Motion
Should be taken care about when pre-process images and conduct statistical analysis. Only motion correction in pre-processing steps is not enough.
- Respiration and heart rate
TR that is too low may give rise to aliasing.
Part 4
相关文章推荐
- DP思想在斐波那契数列递归求解中的应用
- Android MySQL入门
- JQuery图片轮换器原理
- 2015-10-17在北京第四届Elasticsearch中国大会--资料整理
- Android RecyclerView滚动定位
- jquery基本选择器
- LED流水灯(二)
- MySql连接及数据的传输例子
- 原码,补码,反码的由来与转换
- Xcode修改字体设置
- Linux vi命令用法
- POJ 1703 Find them, Catch them(并查集,等价关系)
- HTTP小爬虫 ,nodejs学习(二)
- 3、Spring注解用法的一般步骤
- xcode于Archive当产生安装包遇到ld: library not found for -lPods
- Android 即时通讯
- 连续邮资问题
- oracle数据库优化辅助SQL语句
- Mybatis 3.3.0 Log4j配置
- 10.11 pod 安装