2015 認知神經科學巡迴工作坊 FMRI的實驗設計 講師：張智宏 副教授 中央大學認知神經科學研究所.

Presentation on theme: "2015 認知神經科學巡迴工作坊 FMRI的實驗設計 講師：張智宏 副教授 中央大學認知神經科學研究所."— Presentation transcript:

1 2015 認知神經科學巡迴工作坊 FMRI的實驗設計 講師：張智宏 副教授 中央大學認知神經科學研究所

2 FMRI實驗現場

3 Chouinard, P. A. , Large, M. -E. , Chang, E. C. , & Goodale, M. A
Chouinard, P. A., Large, M.-E., Chang, E. C., & Goodale, M. A. (2009). NeuroImage

4 可能想使用腦造影工具的理由 我想看高僧入定時的腦部活動和一般人有什麼不一樣？

5 使用腦造影技術前，提醒自己： 「如果腦造影結果是答案，那麼研究的問題是什麼？」 「貴重儀器無助於設計不良的實驗。」
2014/1/19 使用腦造影技術前，提醒自己： 「如果腦造影結果是答案，那麼研究的問題是什麼？」 Stephen Kosslyn (1999) 「貴重儀器無助於設計不良的實驗。」 Louis Sokoloff 對實驗結果有合乎理論脈絡的預期，能回達理論上有趣、尚未解決的問題；避免進行釣魚式的腦造影實驗。 能掌握實驗設計以及資料分析概念的研究者，比較有可能生產出有意義的實驗。 這個問題一定要用FMRI來問嗎？

6 實驗基本要素 獨變項(Independent variable, IV) 依變項(Dependent variable, DV)
Aspects of the experimental design that are intentionally manipulated and that are hypothesized to cause changes in DV Conditions or levels At least two conditions/levels for an IV 依變項(Dependent variable, DV) Quantities that are measured to evaluate the effect of IV RT, accuracy, trajectory, … etc. ERP, fMRI, MEG

7 FMRI實驗的基本目的

8 FMRI 實驗術語 Conditions Trials Events

9 設計的概念與方法層面 Conceptual design Methodological design
How to design proper tasks to measure the mental process of interest? Methodological design How to construct task paradigms to optimize the efficiency and power to measure the effects of interest, given multiple constraints in FMRI environment?

10 設計之概念層面 Categorical designs Parametric designs Subtraction Conjunction
Pure insertion, evoked / differential responses Conjunction Testing multiple hypotheses Parametric designs Linear Adaptation, cognitive dimensions Nonlinear Polynomial expansions, neurometric functions

11 Categorical Design Aim: Procedure
Neuronal structures underlying a single process P? Procedure Contrast [Task with P] – [control task without P ] = P The critical assumption of „pure insertion“

12 Example: Cognitive subtraction [Task with P] – [task without P ] = P
2014/1/19 Example: Cognitive subtraction [Task with P] – [task without P ] = P

13 Subtraction Logic: Brain Imaging Example
2014/1/19 Subtraction Logic: Brain Imaging Example Hypothesis : Some areas of the brain are specialized for perceiving objects Simplest design: Compare pictures of objects vs. a control stimulus that is not an object seeing pictures like seeing pictures like minus = object perception Malach et al., 1995, PNAS

14 Objects > Textures Lateral Occipital Complex (LOC)
2014/1/19 Objects > Textures Lateral Occipital Complex (LOC) Malach et al., 1995, PNAS

15 2014/1/19 FMRI Subtraction - =

16 Parametric Design Employs continuous variation in a stimulus/task parameter working memory load, stimulus contrast Inference: Modulation of activity reflects sensitivity to the modulated parameter Can demonstrate more than “where is the activation”: instead, how does this region compute variable X May make control task unnecessary

17 Parametric Design Possible tests for such parametric relation Linear
Nonlinear: Quadratic/cubic/etc. (polynomial expansion) Model-based (e.g. predictions from learning models) IV Level BOLD Change

18 Model-based FMRI

19 2014/1/19 Boynton et al. (1996)

20 Methodological Designs
2014/1/19 Methodological Designs Blocked designs Event-related designs

21 Detection vs. Estimation
2014/1/19 Detection: determination of whether activity of a given voxel (or region) changes in response to the experimental manipulation “which voxel?” Jody 1 Estimation: measurement of the time course within an active voxel in response to the experimental manipulation “How does signal change in a voxel?” % Signal Change 4 8 12 Time (sec) Definitions modified from: Huettel, Song & McCarthy, 2004, Functional Magnetic Resonance Imaging

22 Design Types Block Design Slow ER Design Rapid Counterbalanced
= trial of one type (e.g., face image) = trial of another type (e.g., place image) 2014/1/19 = null trial (nothing happens) Block Design Slow ER Design Rapid Counterbalanced ER Design Jody Rapid Jittered ER Design Mixed Design

23 Block Designs B1 B2 Alternating Design Interleaving null-task blocks

24 2014/1/19 Block Designs = trial of one type (e.g., face image) = trial of another type (e.g., place image) Block Design Early Assumption: Because the hemodynamic response delays and blurs the response to activation, the temporal resolution of fMRI is limited. Jody Positive BOLD response Initial Dip Overshoot Post-stimulus Undershoot 1 2 3 BOLD Response (% signal change) Time Stimulus

25 First fMRI Results with a Block Design
Kwong et al. (1992) PNAS

26 Advantages and Disadvantages
2014/1/19 High detection power Trade-off of block length Long block Larger differences between conditions Short block Avoid confounding with low frequency scanner drift or subject state (like being bored)

27

28 Blocked Design使用建議 Length of a block
Block length at hemodynamic response duration (10~15 s) Equivalent for conditions or combination of conditions to be compared A - B  T(A) = T(B) A + B – C  T(A) + T(B) = T(C) Evoking the same mental process throughout a block

29 Event-related Designs
Slow ER Design Rapid Counterbalanced ER Design Rapid Jittered ER Design

30 Slow Event-Related Designs
2014/1/19 Slow Event-Related Designs Slow ER Design Jody

31 Periodic (Slow) ER Design
Fixed and long ISI Usually > 15s Each event evokes a complete hemodynamic response, and corresponding BOLD are selectively averaged. Inefficient

32 First fMRI Results with an Event-Related Design
Blamire et al. (1992) PNAS

33 Effects of ISI on ER-FMRI Activation
Effects of interstimulus interval on event-related FMRI activation. As the interval between successive events decreases, the overlap between consecutive hemodynamic responses reduces the variability in the BOLD signal. Subjects per- formed a finger tapping task while watching a flashing visual stimulus. Activations within regions of interest in the visual cortex (A) and motor cortex (B) were meas- ured under a number of different experimental conditions. When there was a long interstimulus interval (ISI) of 20 s and a long stimulus duration (SD) of 20 s, mimick- ing a blocked design, there was clear alternating activation in both regions. However, for short-duration events of 2 s, periodic activation was present at long ISIs of 10 s to 12 s but not at short ISIs. (From Bandettini and Cox, 2000.) Bandettinni & Cox (2000)

34 Optimal Constant ISI Brief (< 2 sec) stimuli:
Source: Bandettini et al., 2000 Optimal Constant ISI Brief (< 2 sec) stimuli: optimal trial spacing = 12 sec For longer stimuli: optimal trial spacing = 8 + 2*stimulus duration sec Effective loss in power of event related design: = -35% i.e., for 6 minutes of block design, run ~9 min ER design Jody

35 實驗設計效能 Efficiency of Design
2014/1/19 實驗設計效能 Efficiency of Design Relative measure of desirability of an estimator or experiment design Proportional to power: higher efficient design more likely detects activations Involves comparisons of potentially infinite possibilities/procedures “Given a particular sort of hypothesis to be tested, and with all the constraints for fMRI, how should I present my stimuli to maximize my effect size?”

36 “Do You Wanna Go Faster?”
2014/1/19 Yes, but we have to test assumptions regarding linearity of BOLD signal first Rapid Counterbalanced ER Design Tzvi Rapid Jittered ER Design Mixed Design

37 BOLD response的線性程度 Linearity: “Do things add up?” red = 2 - 1
2014/1/19 Linearity: “Do things add up?” red = 2 - 1 green = 3 - 2 Tzvi Sync each trial response to start of trial Not quite linear but good enough! Source: Dale & Buckner, 1997

38 Linearity is okay for events every ~4+ s

39 快速隨機變動事件相關設計 Rapid Jittered ER Design
2014/1/19 快速隨機變動事件相關設計 Rapid Jittered ER Design = trial of one type (e.g., face image) = trial of another type (e.g., place image) Jody Rapid Jittered ER Design

40 BOLD Overlap With Regular Trial Spacing
2014/1/19 BOLD Overlap With Regular Trial Spacing Neuronal activity from TWO event types with constant ITI Partial tetanus BOLD activity from two event types Slide from Matt Brown

41 隨機變動下的BOLD Overlap 2014/1/19 Neuronal activity from closely-spaced, jittered events BOLD activity from closely-spaced, jittered events Slide from Matt Brown

42 General Linear Model The model Normal equation
2014/1/19 General Linear Model The model Any 𝞫 satisfies the normal equation minimizes the sum of the squares of residuals (e’e) Normal equation Assuming this is invertible

43 Hypothesis Testing df: Coefficient of Efficiency 2014/1/19
Any 𝞫 satisfies the normal equation minimizes the sum of the squares of residuals (e’e) df: Coefficient of Efficiency

44 Why Is Jittered ISI More Efficient?

45 為什麼要變動 ISI？ 增加BOLD signal中可被實驗解釋的變異程度 When pink is on, yellow is off  pink and yellow are anticorrelated Includes cases when both pink and yellow are off  less anticorrelation Without jittering predictors from different trial types are strongly anticorrelated As we know, the GLM doesn’t do so well when predictors are correlated (or anticorrelated)

46 Algorithms for Picking Efficient Designs Optseq2
2014/1/19 Algorithms for Picking Efficient Designs Optseq2

47 Algorithms for Picking Efficient Designs Genetic Algorithms
2014/1/19 Algorithms for Picking Efficient Designs Genetic Algorithms

48 FMRI實驗設計實用建議 作業要引發與研究問題相關之心智歷程 最大化從每位受試者收集的資料量 儘可能多收受試者 在受試者能忍受的前提下
在計畫經費允許的前提下

49 FMRI實驗設計實用建議 選擇可以讓心智歷程產生最大改變的作業情況與時序安排 儘量減少事件之間BOLD的相關性
Jitter or slow ER 建立作業行為表現與BOLD訊號間的相關性

50 比較好的使用腦造影工具的理由實例 高僧因為長期練習入定，其注意力控制能力可能優於常人。預期其注意力網路活化程度，比一般人更高、區域更集中、區域間連結更緊密。 某種心理歷程可能之神經網路為A+B+C…，預期可利用作業TA, TB, TC, TD之間的對比，分別辨識出A, B, and C.

51 參考資料來源：線上 Duke BIAC Dr. Jody Cuhlam’s fMRI for newbies
Dr. Jody Cuhlam’s fMRI for newbies U of Michigan fMRI training course

52 教科書推薦 Huettel et al. (2014) Poldrack et al. (2012)
Linquist & Wager (2015)

53 Questions?

54 But…

55 Other Differences Is subtraction logic valid here?
2014/1/19 Other Differences Is subtraction logic valid here? What else could differ between objects and textures? Objects > Textures object shapes irregular shapes familiarity namability visual features (e.g., brightness, contrast, etc.) actability attention-grabbing Source: Dr. Jody Culham’s fMRI for newbies

56 Other Subtractions > > > Lateral Occipital Complex
2014/1/19 Other Subtractions Lateral Occipital Complex Grill-Spector et al., 1998, Neuron Visual Cortex (V1) > > Kourtzi & Kanwisher, 2000, J Neurosci > Malach et al., 1995, PNAS Source: Dr. Jody Culham’s fMRI for newbies

57 Linearity is okay for events every ~4+ s