Aoioegnigza3l8s0galabeonlneatmtpzAiheabrayeomnlnE大le REVIEW Functional Neuroanatomy of Emotion:A Meta-Analysis of Emotion Activation Studies in PET and fMRI1 K.Luan Phan.Tor Wager,t Stephan F.Taylor.and Israel Liberzon Received November 2.2001 studies with ositron to olved in brain regions are ography PE ic re Sclence (USA) ging (IMK to d va in task dir INTRODUCTION e(s)o and are I It has long been proposed that emotion involves the ht to determin (Pa pez,1937:Mac exis acro ntly.thi en t vielding 761 emission tomography PE nd functional magnetic which in stigated in healthy cts resonance imaging (fMRD).These studies have re Peak acti trans】 ported emot on-relat increase in cerebral blo s h d each r gion that specific brain regions have specialized functions by disgust),to 0 ar-r different induction methods(visual, d tha and In 10 amygdala activations o ond to dis affec the foll (L)The me tive style(Irwin and Davidson 1999) had corte prefronta al role dala:()sad ed with act ty in the and e .2000.wh the orbital prefrontal cortex is considered important emoti nd action by for the related rein ement e o pital he 19989tio retrosplenial rte ited the anterior ate ar sula tasks with ally salient stimuli,particularly in the interaction be cogn Ive de ved th tween emotion and episodic memory (Maddock,1999). critical ce gener greer ab some or thes sing different induction This arch ed in e APIRE/Ja methods and imaging techniques. Severe part b aken separately,individual im ence rd( al I naging studies ca on due to lou arch fel hip T.W.) nd her ity in task design.ima P
REVIEW Functional Neuroanatomy of Emotion: A Meta-Analysis of Emotion Activation Studies in PET and fMRI1 K. Luan Phan,*,2 Tor Wager,† Stephan F. Taylor,* and Israel Liberzon*, ‡ *Department of Psychiatry and †Department of Psychology, University of Michigan, Ann Arbor, Michigan 48109; and ‡Psychiatry Service, Ann Arbor VAMC, Ann Arbor, Michigan 48105 Received November 2, 2001 Neuroimaging studies with positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) have begun to describe the functional neuroanatomy of emotion. Taken separately, specific studies vary in task dimensions and in type(s) of emotion studied and are limited by statistical power and sensitivity. By examining findings across studies, we sought to determine if common or segregated patterns of activations exist across various emotional tasks. We reviewed 55 PET and fMRI activation studies (yielding 761 individual peaks) which investigated emotion in healthy subjects. Peak activation coordinates were transformed into a standard space and plotted onto canonical 3-D brain renderings. We divided the brain into 20 nonoverlapping regions, and characterized each region by its responsiveness across individual emotions (positive, negative, happiness, fear, anger, sadness, disgust), to different induction methods (visual, auditory, recall/imagery), and in emotional tasks with and without cognitive demand. Our review yielded the following summary observations: (1) The medial prefrontal cortex had a general role in emotional processing; (2) fear specifically engaged the amygdala; (3) sadness was associated with activity in the subcallosal cingulate; (4) emotional induction by visual stimuli activated the occipital cortex and the amygdala; (5) induction by emotional recall/imagery recruited the anterior cingulate and insula; (6) emotional tasks with cognitive demand also involved the anterior cingulate and insula. This review provides a critical comparison of findings across individual studies and suggests that separate brain regions are involved in different aspects of emotion. © 2002 Elsevier Science (USA) INTRODUCTION It has long been proposed that emotion involves the limbic system (Papez, 1937; MacLean, 1952; LeDoux, 1996). Recently, this assumption has been tested with functional neuroimaging techniques such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). These studies have reported emotion-related increases in cerebral blood flow or BOLD signal (activations) in cortical, limbic, and paralimbic regions. Many authors have hypothesized that specific brain regions have specialized functions for emotional operations. For example, while some postulated that the amygdala is critical to fear-related processing (LeDoux, 2000), others have suggested that amygdala activations correspond to dispositional affective style (Irwin and Davidson, 1999). The medial prefrontal cortex has been hypothesized to have specific roles for emotional decision making (Damasio, 1996) and emotional self-regulation (Davidson, 2000), while the orbital prefrontal cortex is considered important for the evaluation of emotion-related reinforcement contingencies (Rolls, 1999). The retrosplenial cortex has been proposed as important in processing emotionally salient stimuli, particularly in the interaction between emotion and episodic memory (Maddock, 1999). In spite of general agreement about some of these specialized emotional regions, conflicting findings are often produced by studies using different induction methods and imaging techniques. Taken separately, individual imaging studies cannot fully characterize which brain regions are responsible for emotion due to low statistical power and heterogeneity in task design, imaging methods, and analysis. These variations have made it difficult to interpret the 1 This research was supported in part by the APIRE/Janssen Research on Severe Mental Illness Award (K.L.P.), the Rachel Upjohn Clinical Neuroscience Scholars Award (K.L.P.), the Mental Illness Research Association (K.L.P.), and the National Science Foundation Graduate Research Fellowship (T.W.). 2 To whom correspondence and reprint requests should be addressed. Fax: (734) 647-8514. E-mail: luan@umich.edu. NeuroImage 16, 331–348 (2002) doi:10.1006/nimg.2002.1087, available online at http://www.idealibrary.com on 331 1053-8119/02 $35.00 © 2002 Elsevier Science (USA) All rights reserved
332 PHAN ET AL. s found.Unde this et al.1998b:LaBar et al.1998:Buchel et al.1999 s I ve udies one a 21100g B iti n and ex across studies patterns of activa nde them ine hle to the tions can be evaluated across similar and dissimilar tional tasks in our database.Furthermore.ths fe emotional tasks.This meta-analysis examines findings conditioning related activations were discussed in a recent review by Buchel and Dolan (2000).0n ific activation peaks were examined in this md if there are meta-analysis.The reporting of deactivation or de with emotional activation tasks that had a comitive creases in din acti alty was en component (e.g.. emotional ni tod mined activations remain undetermine and their spe in regions w ere to ifferent tio s re ain inconclusive or unclear (Hutchinson a em 1999:Raichle et al 2001) ditio cer tain brain regions were for different emotional responses Organization of Results (reflected by the frequency of activation in a s Fifty-five publications/studies (43 PET and 12 fMRI) region ith tas ning from May.1993.to December.2000.that met T ehtfanteria.yalding19subtractiornscoatst atterns and the and 761 individual activation peaks,were included for stion of laterality in activation patterns are major topics in neuroimagi meta-analysis(Table 1).Because the studies adopted dif ng of emotion and thus require an extensive and separate discussion. all foci the Therefore,we have as sig ficant by hosen to report these results sep- ia d lual stu arately (Wager et nanner:(1)Res ith Indivdual Emotion METHODS Socated with Induction Method (visual auditpry.auto happiness):(2)regions as Scope of Review plographical recall/imagery): and (3)regions associated with preser e and abs ence of Cognitive Demand.Table sof activa on uy,an a large for English-language manuscripts of PET and fMRI he af rate but emotionindction5udespubihedbetwenpdaotm2a nent of Induction Method.for a variety of reaso 1990,and Decemb To allow us to pertorm literature often distinguishes between er,200 11T Ireports otional tasks contain varous degrees of cognitive de. s②Th mand.Furthermore,there is a clear interaction between tal processes of emotion (thus.studies of lower -order lon and cog onon a eve To exa sensory or motor processes,such as gustatory/olfactory we ti a were excluded)see reviews er al.1999 vation b s in which an emotional task was co pled with a concurren nitive task (e.g en BOLD-MRD der/emotional expression discrimination.emotional rat- across the entire brain (i.e..excl studi tha ing,picture/face recognition/encoding.naming,counting focused on limited regions of the brain):(4)They all autobiographical recall/imagery.etc.)as Emotion- Cog used the image subtraction methodology to determine or eman provided ta dard a irach n the em rnou Montreal pa (MNI ith nitive Demand This classification allo ws us to examing laboratories.We chose not to include studies on aver the effect of a "nonemotional" ognitive component on sive and trace conditioning (Buchel et al,1998:Morris emotional tasks
differences in activation patterns found. Under this circumstance, a broader-based meta-analysis of multiple studies may be one solution (Fox et al., 1998). By examining findings across studies, patterns of activations can be evaluated across similar and dissimilar emotional tasks. This meta-analysis examines findings across imaging studies in search of specific regions associated with emotional activation in general, with specific emotions and different induction methods. We also examined if there are brain regions associated with emotional activation tasks that had a cognitive component (e.g., emotional expression recognition, gender discrimination, etc.). Particularly, we examined how “sensitive” specific brain regions were to different emotional tasks (reflected by the percentage of studies reporting activation in a region according to a condition of interest). We also examined how “specific” certain brain regions were for different emotional responses (reflected by the frequency of activation in a specific region with a given task in comparison to other regions). The effect of gender and valence (the extent to which emotion is unpleasant or pleasant) on activation patterns, and the question of laterality in activation patterns are major topics in neuroimaging of emotion, and thus require an extensive and separate discussion. Therefore, we have chosen to report these results separately (Wager et al., in preparation). METHODS Scope of Review In order to illuminate both general and specific patterns of activation associated with different emotional tasks, we searched peer-reviewed journals (indexed in large databases [MEDLINE, PsychInfo, BrainMap]) for English-language manuscripts of PET and fMRI emotion induction studies published between January, 1990, and December, 2000. To allow us to performed planned meta-analysis, all reports included met the following criteria: (1) They involved unmedicated healthy adults; (2) They focused on higher-order mental processes of emotion (thus, studies of lower-order sensory or motor processes, such as gustatory/olfactory or pain induction, were excluded) [see reviews by by Small et al., 1999; Casey et al., 1994, respectively]; (3) They all measured regional cerebral blood flow (e.g., O15H2O-PET) or blood oxygenation (e.g., BOLD-fMRI) across the entire brain (i.e., excluding studies that focused on limited regions of the brain); (4) They all used the image subtraction methodology to determine activation foci; (5) They provided standard Talairach (Talairach and Tournoux, 1988) or Montreal Neurologic Institute (MNI) coordinates, allowing for comparison of findings across different studies and different laboratories. We chose not to include studies on aversive and trace conditioning (Bu¨ chel et al., 1998; Morris et al., 1998b; LaBar et al., 1998; Bu¨ chel et al., 1999) because those tasks involve associative learning and behavioral conditioning (including acquisition and extinction), rendering them incomparable to the emotional tasks in our database. Furthermore, these fearconditioning related activations were extensively discussed in a recent review by Bu¨ chel and Dolan (2000). Only activation peaks were examined in this meta-analysis. The reporting of deactivation or decreases in brain activity was not consistent across studies which did not allow meaningful generalization. Also, the neural mechanisms underlying reported deactivations remain undetermined and their interpretations remain inconclusive or unclear (Hutchinson et al., 1999; Raichle et al., 2001). Organization of Results Fifty-five publications/studies (43 PET and 12 fMRI) spanning from May, 1993, to December, 2000, that met our database criteria, yielding 119 subtractions/contrasts and 761 individual activation peaks, were included for meta-analysis (Table 1). Because the studies adopted different analysis methods and significance criteria, all foci were accepted when reported as significant by the criteria designated in the individual studies. The activation results are grouped in the following manner: (1) Regions associated with Individual Emotion (fear, sadness, disgust, anger, happiness); (2) regions associated with Induction Method (visual, auditory, autobiographical recall/imagery); and (3) regions associated with presence and absence of Cognitive Demand. Table 1 lists all studies included in the review, arranged alphabetically, and identifies the Individual Emotion examined and the Induction Method employed. We examined the effect of Cognitive Demand, as a separate but related component of Induction Method, for a variety of reasons. Neuroimaging literature often distinguishes between cognitive and emotional tasks, but the majority of the emotional tasks contain various degrees of cognitive demand. Furthermore, there is a clear interaction between emotion and cognition on a functional level. To examine the neuroanatomic basis of this interaction, we examined the effect of these cognitive components in emotion activation by grouping conditions in which an emotional task was coupled with a concurrent cognitive task (e.g., gender/emotional expression discrimination, emotional rating, picture/face recognition/encoding, naming, counting, autobiographical recall/imagery, etc.) as Emotion Cognition or with Cognitive Demand. Conversely, we grouped conditions in which the emotional task did not explicitly have a cognitive component (i.e., passive viewing, passive listening) as Emotion alone or without Cognitive Demand. This classification allows us to examine the effect of a “nonemotional” cognitive component on emotional tasks. 332 PHAN ET AL
FUNCTIONAL NEUROANATOMY OF EMOTION 333 TABLE 1 List of Emotion Activation Studies Included in the Meta-Analysis Induction Methoc Type of Emotio Study No. Reference Visual Auditory Recall Happy Fear Anger Sad Disgust 9 d97 ard 9 d99 nli 98 y99 199 111213145677809001223450667896006123456008890426456469505268 Har 19 L 99 is neye ”98 55 Whalen 98b The stand coor tes of activation peaks re- 96.Wellcome Department of Cognitive Neurology,Lon and medial i onto I SPM IA-IC show the es va
The standard coordinates of activation peaks reported by individual studies were plotted onto lateral and medial views of a 3-D canonical brain image (SPM 96, Wellcome Department of Cognitive Neurology, London; derived from the MNI brain template). Figures 1A–1C show the result of grouping plotted activaTABLE 1 List of Emotion Activation Studies Included in the Meta-Analysis Study No. Reference Induction Method Type of Emotion Visual Auditory Recall Happy Fear Anger Sad Disgust 1 Baker 97 X X X 2 Beauregard 97 X 3 Beauregard 98 X X 4 Blair 99 X X X 5 Blood 99 X 6 Breiter 96 X X X 7 Canli 98 X 8 Crosson 99 X 9 Damasio 00 X X X X X 10 Dolan 00 X 11 Dougherty 99 X X 12 Frey 00 X 13 Gemar 96 X X 14 George 93 X 15 George 94 X X 16 George 95 X X X 17 George 96a X 18 George 96b X X X 19 Hamann 99 X 20 Hariri 00 X 21 Isenberg 99 X X 22 Kimbrell 99 X X X 23 Kosslyn 96 X 24 Lane 97a X 25 Lane 97b X 26 Lane 97c X X X X X 27 Lane 98 X X 28 Lane 99 X 29 Liberzon 00 X 30 Liotti 00 X X X 31 Maddock 97 X X 32 Mayberg 99 X X 33 Morris 96 X X X 34 Morris 98a X X X 35 Morris 99 X X 36 Nakamura 99 X 37 Paradiso 97 X X X X 38 Paradiso 99 X 39 Pardo 93 X X 40 Partiot 95 X X 41 Phillips 97 X X X 42 Phillips 98a X X X X 43 Phillips 98b X X X 44 Pietrini 00 X 45 Rauch 99 X 46 Redoute 00 X 47 Reiman 97 X X 48 Royet 00 X X 49 Shin 00 X 50 Simpson 00 X 51 Sprengelmeyer 98 X X X X 52 Taylor 98 X 53 Taylor 00 X 54 Teasdale 99 X 55 Whalen 98b X FUNCTIONAL NEUROANATOMY OF EMOTION 333
334 PHAN ET AL. Visual Auditory Recall
FIG. 1A. Activation foci: Individual emotion. FIG. 1B. Activation foci: Induction method. 334 PHAN ET AL
FUNCTIONAL NEUROANATOMY OF EMOTION 36 ☐Emotion+cognition Emotion alone FIG.IC. Activ n foci:Cognitive demand tage of studies that reported an acti- specific tion in se to each rect con parisons across studies.we translated re- Indivumoton.Indun Methd.no (alarachand e nttp: .cam.ac can lead to D os that we chose not to include spatial extent of activation compared the number of studies that found activation peaks when plotting onto the 3-D canonical brain. in a particular region to those that did not using chi- For a semiquantitative analysis,we divided the atl square (X)analysis.The results of the X?analysis are into an presented in Fig.2. egowaoas0atedy udy included in this review RESULTS AND DISCUSSION tomical structure/gyrus and/or Brodmann area. We 1.Regions Involved Across Individual Emotions on to l lize the activation pea s in rCereral emgtionali and the medial 0 tudy's cho hr shold d sidered a re that no single brain egion is commonly activated bya more over 50%of all stuc media proach was choser to counterba ency fo cortex (MPFC)was commonly activated natiionmnetiod2tepig 2A and 2B)
tion foci according to Individual Emotion, Induction Method, and Cognitive Demand. In order to make direct comparisons across studies, we translated reported Talairach coordinates (Talairach and Tournoux, 1998) into MNI coordinates (transformation developed by Matthew Brett, http://www.mrc.cbu.cam.ac.uk/Imaging). Because differences in image smoothing techniques can lead to different numbers of activation foci, we chose not to include spatial extent of activation peaks when plotting onto the 3-D canonical brain. For a semiquantitative analysis, we divided the atlas brain into 20 general regions, and examined whether activations in each specific region was associated with different Individual Emotion, Induction Method, and Cognitive Demand. Each study included in this review identified the location of the activation peak as anatomical structure/gyrus and/or Brodmann area. We used this information to localize the activation peaks in the 20 brain regions used in this review. The number of activation peaks reported for a single region differed according to each study’s chosen statistical threshold and analysis methods. Therefore, we considered a region as activated for a particular study if one or more activation peak in this region was reported. This approach was chosen to counterbalance the tendency for overestimating activations based on variable thresholds used in different studies, and allowed us to estimate the percentage of studies that reported an activation foci in a specific region in response to each Individual Emotion, Induction Method, and Cognitive Demand (Figs. 2A–2C). Additionally, we examined how specific the reported regional activations were to Individual Emotion, Induction Method, and Cognitive Demand. For all studies that employed similar contrasts and methods, we compared the number of studies that found activation in a particular region to those that did not using chisquare (X2 ) analysis. The results of the X2 analysis are presented in Fig. 2. RESULTS AND DISCUSSION 1. Regions Involved Across Individual Emotions 1.1. General emotional processing and the medial prefrontal cortex. No specific brain region was consistently activated in the majority of studies, across individual emotions and induction methods, suggesting that no single brain region is commonly activated by all emotional tasks. Although no region was activated in over 50% of all studies, we did find that the medial prefrontal cortex (MPFC) was commonly activated, and that its activation was not specific to a specific emotion or induction method (see Figs. 2A and 2B). FIG. 1C. Activation foci: Cognitive demand. FUNCTIONAL NEUROANATOMY OF EMOTION 335