## Spatial Pattern Analysis of Functional Brain Images Using Partial Least Squares (1996)

Venue: | Neuroimage |

Citations: | 87 - 11 self |

### BibTeX

@ARTICLE{Mcintosh96spatialpattern,

author = {A. R. Mcintosh and F.L. Bookstein and J. V. Haxby and C. L. Grady},

title = {Spatial Pattern Analysis of Functional Brain Images Using Partial Least Squares},

journal = {Neuroimage},

year = {1996},

volume = {3},

pages = {143--157}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper introduces a new tool for functional neuroimage analysis: partial least squares (PLS). It is unique as a multivariate method in its choice of emphasis for analysis, that being the covariance between brain images and exogenous blocks representing either the experiment design or some behavioral measure. Whatemerges are spatial patterns of brain activity that represent the optimal association between the images and either of the blocks. This process differs substantially from other multivariate methods in that rather than attempting to predict the individual values of the image pixels, PLS attempts to explain the relation between image pixels and task or behavior. Data from a face encoding and recognition PET rCBF study are used to illustrate two types of PLS analysis: an activation analysis of task with images and a brain-- behavior analysis. The commonalities across the two analyses are suggestive of a general face memory network differentially engaged during encoding and recognition. PLS thus serves as an important extension by extracting new information from imaging data that is not accessible through other currently used univariate and multivariate image analysis tools. r 1996 Academic Press, Inc

### Citations

1626 | Co-planar stereotaxic atlas of the human brain. 3-dimensional proportional system: an approach to cerebral imaging - Talairach, Tournoux - 1988 |

924 |
Multivariate analysis
- Mardia, Kent, et al.
- 1979
(Show Context)
Citation Context ...trix of order n of which the vector B of the preceding discussion is the first column, and D is an m 3 n diagonal matrix of which the scalar d of the preceding discussion is the upper left entry (see =-=Mardia et al., 1979-=-). This can be written more explicitly as the reconstruction S 5 ADBt t t 5 d1A1DB1 1 d2A2B2 1 .... The interpretation in terms of explained sums-ofsquares of the matrix S begins with the rank-1 appro... |

225 |
Bootstrap methods for standard errors, confidence intervals, and other measures of statistical accuracy
- Efron, Tibshirani
- 1986
(Show Context)
Citation Context ...od can be employed for the saliences themselves. In this case, bootstrap estimates of standard errors provide a relatively simple method to determine the precision of the saliences within a given LV (=-=Efron and Tibshirani, 1986; se-=-e also Sampson et al., 1989 for an application in PLS). In this paper, we simply interpret the saliences as images of spatial pattern—the apparent concentrations in Figs. 2–4 are a guide to neurop... |

207 | Encoding specificity and retrieval processes in episodic memory - Tulving, Thompson - 1973 |

198 |
Morphometric tools for landmark data: geometry and biology
- Bookstein
- 1991
(Show Context)
Citation Context ...m of the correlations of RT with rCBF in the two tasks separately, and the second column is the difference of those correlations. Because PLS is invariant against rotation of either block separately (=-=Bookstein, 1991-=-), this could as well have been represented as the unrotated matrix of the two task-specific correlations separately, followed by a rotation in the course of reporting the results.s6. RESULTS SPM Resu... |

101 | Dynamics of neuronal firing correlation: Modulation of “effective connectivity
- Aertsen, Gerstein, et al.
- 1989
(Show Context)
Citation Context ...ffering degrees. For instance, methods introduced to neuroimaging by McIntosh and GonzalezLima (1994) for network analysis and by Friston (1994) for measures of functional and effective connectivity (=-=Aertsen et al., 1989-=-; Gerstein et al., 1978) represent the explicit use of interdependency as indices of neural interactions. This paper calls attention to a newer method originally developed for applications to one-dime... |

98 |
Randomization tests
- Edgington
(Show Context)
Citation Context ...he quantities that characterized the computation—the saliences of either set of singular vectors or the scores that characterize the experimental tasks. In this application we use a permutation test=-= (Edgington, 1980-=-; Good, 1994) to assess the extent to which the scores characterize tasks. The test involves randomly reordering the rows of the design matrix (i.e., destroying the association of experimental conditi... |

87 |
Nonparametric analysis of statistic images from functional mapping experiments
- Holmes, Blair, et al.
- 1996
(Show Context)
Citation Context ...t relying on the distributional assumptions underlying most conventional parametric statistical methods. A univariate version of permutation tests for functional mapping has recently been introduced (=-=Holmes et al., 1996-=-). The test used in this application was a multiple regression approach to ANOVA (Pedhazur, 1982). The collected condition contrast vectors were used as predictor variables and the scores from each si... |

81 |
Structural equation modelling and its application to network analysis in function brain mapping
- McIntosh, Gonzalez-Lima
- 1994
(Show Context)
Citation Context ..., associations of the image with its causes or effects are treated as a whole, consistent with our scientific stance of regarding brain operations as outcomes of cooperative interactions among areas (=-=McIntosh and Gonzalez-Lima, 1994-=-). At the same time, the permutation tests underlying significance here avoid problems of multiple comparisons that bedevil other approaches to the image as a whole such as the eigenimage analysis. To... |

67 |
Functional connectivity: the principal-component analysis of large (PET) data sets
- Friston, Frith, et al.
- 1993
(Show Context)
Citation Context ...e descriptors, the findings of PLS are carried by singular images, computed multivariate optima somewhat related to the previously introduced subscale profiles (Moeller et al., 1987) and eigenimages (=-=Friston et al., 1993-=-). The singular images are spatial patterns of pixel contents representing pixelby-pixel covariances with task or behavior. For a thorough review of PLS, including a tour of its scientific antecedents... |

54 | Characterizing dynamic brain responses with fMRI—A multivariate approach - Friston, Frith, et al. - 1995 |

43 |
Functional brain maps of retrieval mode and recovery of episodic information
- Nyberg, Tulving, et al.
- 1995
(Show Context)
Citation Context ...There have been suggestions that recognition of previously presented information requires reactivation of some of the same regions engaged in the initial encoding episode (Tulving and Thompson, 1973; =-=Nyberg et al., 1995-=-). The PLS results are consistent with this possibility. TABLE 3 Local Maxima in Areas Identified from PLS Analysis of Behavior/Brain Relations Location X Y Z SI Thalamus 22 22 12 SI11 Medial occipita... |

41 |
Permutation Tests
- Good
- 1994
(Show Context)
Citation Context ...t characterized the computation—the saliences of either set of singular vectors or the scores that characterize the experimental tasks. In this application we use a permutation test (Edgington, 1980=-=; Good, 1994-=-) to assess the extent to which the scores characterize tasks. The test involves randomly reordering the rows of the design matrix (i.e., destroying the association of experimental conditions with ima... |

41 | Network analysis of cortical visual pathways mapped with PET - McIntosh, Grady, et al. - 1994 |

29 | Age-related reductions in human recognition memory due to impaired encoding - Grady, McIntosh, et al. - 1995 |

22 |
A regional covariance approach to the analysis of functional patterns in positron emission tomographic data
- Moeller, Strother
- 1991
(Show Context)
Citation Context ...er image analysis methods have been proposed that attempt to extract images related to either experimental manipulations or experimental groups. Scaled subprofile modeling (SSM; Moeller et al., 1987; =-=Moeller and Strother, 1991-=-; Alexander and Moeller, 1994) performs a principal components analysis on a covariance matrix obtained between brain regions across a sample of subjects containing both normal controls and patients. ... |

18 |
Functional and effective connectivity: A synthesis
- Friston
- 1994
(Show Context)
Citation Context ... Although they are not formally optimal for any discriminatory purpose, the eigenimages can be powerful aids in description of group differences. The eigenimage analysis of Friston et al. (1993; also =-=Friston, 1994-=-) differs from that of SSM in that subject variance is completely removed by averaging and the analysis is performed across tasks. The goal of the analysis here is to identify those brain regions most... |

15 |
Scaled Subprofile Model: A Statistical Approach to the Analysis
- Moeller, Strother
(Show Context)
Citation Context ... correlations. Within the space of image descriptors, the findings of PLS are carried by singular images, computed multivariate optima somewhat related to the previously introduced subscale profiles (=-=Moeller et al., 1987-=-) and eigenimages (Friston et al., 1993). The singular images are spatial patterns of pixel contents representing pixelby-pixel covariances with task or behavior. For a thorough review of PLS, includi... |

13 | Plastic transformation of PET images - Friston, Frith, et al. - 1991 |

13 |
Identification of functionally related neural assemblies
- Gerstein, Perkel, et al.
- 1978
(Show Context)
Citation Context ...nstance, methods introduced to neuroimaging by McIntosh and GonzalezLima (1994) for network analysis and by Friston (1994) for measures of functional and effective connectivity (Aertsen et al., 1989; =-=Gerstein et al., 1978-=-) represent the explicit use of interdependency as indices of neural interactions. This paper calls attention to a newer method originally developed for applications to one-dimensional images from spe... |

11 |
The Enduring Effects of Prenatal Alcohol Exposure on Child Development, Birth through Seven Years: A Partial Least Squares Solution
- Streissguth, Bookstein, et al.
- 1993
(Show Context)
Citation Context ...ition, not the eigenanalysis of the Y-block per se, that a threshold of significance is appropriately imposed for relevance of image contents to the experimental design or to behavioral measures (see =-=Streissguth et al., 1993-=-, for a two-dimensional example). The number of terms that prove significant in this decomposition typically is small. A PLS analysis thus combines the elements of two (or more) blocks of variables in... |

10 |
Face encoding and recognition in the human
- Haxby, Ungerleider, et al.
- 1996
(Show Context)
Citation Context ...ntrations in Figs. 2–4 are a guide to neurophysiological interpretation. 4. PET DATA The data were obtained from a PET rCBF study that compared encoding and recognition of faces in 10 young subjects=-= (Haxby et al., 1996-=-). The experiment had four conditions: (1) simultaneous match-to-sample of faces (face matching), (2) encoding of faces, (3) two-alternative recognition of faces, and (4) a sensorimotor control task. ... |

9 |
Multiple regression in behavioural research: Explanation and prediction
- Pedhazur
- 1997
(Show Context)
Citation Context ...ds. A univariate version of permutation tests for functional mapping has recently been introduced (Holmes et al., 1996). The test used in this application was a multiple regression approach to ANOVA (=-=Pedhazur, 1982-=-). The collected condition contrast vectors were used as predictor variables and the scores from each singular image were used as dependent variables. The computation of the test statistic is straight... |

5 |
Neurobehavioral effects of prenatal alcohol: Part II. Partial Least Squares analysis. Neurotoxicology and tetralogy
- Sampson, Streissguth, et al.
- 1989
(Show Context)
Citation Context ...themselves. In this case, bootstrap estimates of standard errors provide a relatively simple method to determine the precision of the saliences within a given LV (Efron and Tibshirani, 1986; see also =-=Sampson et al., 1989 for-=- an application in PLS). In this paper, we simply interpret the saliences as images of spatial pattern—the apparent concentrations in Figs. 2–4 are a guide to neurophysiological interpretation. 4.... |

4 |
Application of the Scaled Subprofile Model to Functional Imaging in Neuropsychiatric Disorders: A Principal Component Approach to Modeling Brain Function in Disease
- Alexander, Moeller
- 1994
(Show Context)
Citation Context ...erve to enhance the cross-block prediction of task effects. If they differ, then the data may be adjusted for the effects from subject or run, either by preprocessing steps (cf. Moeller et al., 1987; =-=Alexander and Moeller, 1994-=-) or by adjusting out singular images representing such effects. Three other image analysis methods have been proposed that attempt to extract images related to either experimental manipulations or ex... |

4 | Partial least squares: a dose–response model for measurement in the behavioral and brain sciences,” Psycoloquy 5 - Bookstein - 1994 |

3 |
Biometrics and brain maps: the promise of the Morphometric Synthesis
- Bookstein
- 1996
(Show Context)
Citation Context ...es are crucial to the computations that follow; a careful consideration of rules for their imposition is the subject of another paper. For a very simple example, a single contrast between groups, see =-=Bookstein, 1996-=-.) Then the vector B will always be an image of unit amplitude (sum-of-squares) representing variation of normalized pixels around the grand mean image. Because these images are singular vectors of th... |

1 | Measuring ‘dose’ and ‘response’ with multivariate data using partial least squares techniques - Bookstein, Sampson, et al. - 1990 |

1 | Exploiting redundant measurement of dose and behavioural outcome: New methods from the teratology of alcohol - Bookstein, Streissguth, et al. - 1996 |