Elsevier

NeuroImage

Volume 34, Issue 1, 1 January 2007, Pages 137-143
NeuroImage

Technical Note
Biological parametric mapping: A statistical toolbox for multimodality brain image analysis

https://doi.org/10.1016/j.neuroimage.2006.09.011Get rights and content

Abstract

In recent years, multiple brain MR imaging modalities have emerged; however, analysis methodologies have mainly remained modality-specific. In addition, when comparing across imaging modalities, most researchers have been forced to rely on simple region-of-interest type analyses, which do not allow the voxel-by-voxel comparisons necessary to answer more sophisticated neuroscience questions. To overcome these limitations, we developed a toolbox for multimodal image analysis called biological parametric mapping (BPM), based on a voxel-wise use of the general linear model. The BPM toolbox incorporates information obtained from other modalities as regressors in a voxel-wise analysis, thereby permitting investigation of more sophisticated hypotheses. The BPM toolbox has been developed in Matlab with a user-friendly interface for performing analyses, including voxel-wise multimodal correlation, ANCOVA, and multiple regression. It has a high degree of integration with the SPM (statistical parametric mapping) software relying on it for visualization and statistical inference. Furthermore, statistical inference for a correlation field, rather than a widely used T-field, has been implemented in the correlation analysis for more accurate results. An example with in vivo data is presented, demonstrating the potential of the BPM methodology as a tool for multimodal image analysis.

Introduction

Functional MR imaging (fMRI) has revolutionized the field of neuroscience and has emerged as a widely used research tool for the probing of neural processes. It has evolved predominantly using BOLD (Blood Oxygen Level Dependent) fMRI techniques (Ogawa et al., 1990). Over the last several years, additional imaging modalities have emerged, including diffusion tensor imaging (DTI), perfusion imaging, T2 mapping, 3D spectroscopic imaging, and a variety of anatomic-based methods such as voxel-based morphometry (VBM). There is, however, a relative lack of available multimodal image analysis methodologies. That is, analysis for each form of functional imaging data has mainly proceeded to develop specific to that imaging modality. Despite this, there has been some work in multimodality massively univariate analyses. For example, Richardson et al. (1997) proposed a univariate technique, a voxel-wise group-by-modality interaction analysis based on SPM (statistical parametric mapping) and applied to the analysis of PET and MRI data recorded from normal and epileptic subjects. In another univariate approach (Pell et al., 2004), a conjunction analysis (Friston et al., 1999) was applied to voxel-based morphometry (VBM) and T2-relaxometry data to assess the degree of concurrent changes in both data sets. A non-parametric method has recently been proposed based on the use of combining functions and permutation testing that is able to detect not only areas of concurrent changes but also disassociated changes across modalities (Hayasaka et al., 2006). None of these massively univariate approaches, however, can explicitly model changes in one imaging modality as a function of another modality. Only recently, a data-driven method has been proposed as an exploratory tool for multiple modality imaging based on independent component analysis (ICA) (Calhoun et al., 2006).

In this technical note, we describe an approach to multimodal integrative image analysis using the biological parametric mapping (BPM) software package. The BPM software allows probing of neuroimaging data using information from other functional or structural imaging modalities. The major conceptual difference between a BPM analysis and a conventional SPM style analysis is in the use of biological information, such as choline concentration or tissue anisotropy, obtained from one or more imaging modalities, as regressors in an analysis of another imaging modality in a massively univariate fashion. For example, a BPM analysis can determine if changes in metabolite levels (from 3D spectroscopy) or fractional anisotropy (in diffusion tensor data) correlate with activations in an fMRI experiment and vice versa. The statistical concepts embodied in a BPM-style analysis are not novel and can be found in any general textbook of statistics. However, the application of these methods to neuroimaging research provides a powerful tool for hypothesis testing and discovery that otherwise is not available.

We describe the BPM toolbox that incorporates correlation, ANCOVA, and multiple regression analyses of multimodal imaging data sets. Here, we utilize the toolbox to analyze in vivo data. The BPM toolbox uses the theoretical framework behind the widely used SPM methodology: the general linear model (GLM) (Friston et al., 1995) for statistical estimation and random field theory (RFT) (Worsley et al., 1996) or false discovery rate (FDR) (Benjamini and Hochberg, 1995, Genovese et al., 2002) for statistical inference. In addition to the RFT-based inference on T- and F-images, we have implemented RFT-based inference for correlation images (Cao and Worsley, 1999) in our correlation analysis, which is more appropriate in this context.

Section snippets

Materials and methods

BPM allows solving a GLM with a different design matrix in each voxel (Fig. 1). The difference in design matrices is due to allowing image data covariates. Supported analyses include ANOVA, ANCOVA, correlation, and multiple regression. The BPM toolbox was developed entirely in Matlab (MathWorks; Natick, MA, USA) as an open source package. It was designed to only use routines available in a standard Matlab distribution and does not require any additional Matlab toolboxes. The BPM toolbox has a

Results

Fig. 5 shows the results of the in vivo fMRI ANOVA, VBM ANOVA, and ANCOVA comparing normal and dyslexic subjects (Dyslexic > Normal), as well as the correlation analysis results. The statistical maps were thresholded at p < 0.0025 and corrected for cluster extent at p < 0.05. In the ANOVA performed between the two groups of subjects, a focus of activation is present (Fig. 5A) in the left temporo-occipital area (Brodmann area 37), representing increased activity in the dyslexic readers performing the

Discussion

In this paper, we have described a method for performing SPM group analysis with image data as an independent variable using the BPM software package. It is a massively univariate approach that utilizes the general linear model for statistical estimation and RFT or FDR for statistical inference. Although the BPM package is designed to take imaging data as covariates, it can also accept scalar dependent measures into the design matrix. The use of the SPM inference and visualization tools will

Conclusion

The BPM toolbox is a necessary evolutionary step in the deployment of methods for multimodal functional image data analysis and rapid testing of sophisticated imaging-based hypotheses. By combining information from different imaging modalities, the BPM approach overcomes the limitation of modality-specific analysis inherent to current fMRI software packages. The method, however, should not be conceptually limited to fMRI analyses. This tool could be used to analyze any brain imaging data that

Acknowledgments

This work was supported by the Human Brain Project and NIBIB through grant number EB004673, and in part by NS042568, and P01-HD-21887. We would like to thank Dr. Ann Peiffer and Ms. Christina Hugenschmidt for their extensive beta-testing of the BPM toolbox and Ms. Kathy Pearson for help with computer programming.

References (17)

There are more references available in the full text version of this article.

Cited by (268)

View all citing articles on Scopus
View full text