Model Assessment and Model Building in fMRI

Mehrdad Razavi, Thomas J. Grabowski, Walter P. Vispoel, Patrick Monahan, Sonya Mehta, Brent Eaton, Lizann Bolinger

Research output: Contribution to journalArticle

23 Citations (Scopus)

Abstract

Model quality is rarely assessed in fMRI data analyses and less often reported. This may have contributed to several shortcomings in the current fMRI data analyses, including: (1) Model mis-specification, leading to incorrect inference about the activation-maps, SPM{t} and SPM{F}; (2) Improper model selection based on the number of activated voxels, rather than on model quality; (3) Underutilization of systematic model building, resulting in the common but suboptimal practice of using only a single, pre-specified, usually over-simplified model; (4) Spatially homogenous modeling, neglecting the spatial heterogeneity of fMRI signal fluctuations; and (5) Lack of standards for formal model comparison, contributing to the high variability of fMRI results across studies and centers. To overcome these shortcomings, it is essential to assess and report the quality of the models used in the analysis. In this study, we applied images of the Durbin-Watson statistic (DW-map) and the coefficient of multiple determination (R2-map) as complementary tools to assess the validity as well as goodness of fit, i.e., quality, of models in fMRI data analysis. Higher quality models were built upon reduced models using classic model building. While inclusion of an appropriate variable in the model improved the quality of the model, inclusion of an inappropriate variable, i.e., model mis-specification, adversely affected it. Higher quality models, however, occasionally decreased the number of activated voxels, whereas lower quality or inappropriate models occasionally increased the number of activated voxels, indicating that the conventional approach to fMRI data analysis may yield sub-optimal or incorrect results. We propose that model quality maps become part of a broader package of maps for quality assessment in fMRI, facilitating validation, optimization, and standardization of fMRI result across studies and centers.

Original languageEnglish (US)
Pages (from-to)227-238
Number of pages12
JournalHuman Brain Mapping
Volume20
Issue number4
DOIs
StatePublished - Dec 2003
Externally publishedYes

Fingerprint

Magnetic Resonance Imaging

Keywords

  • Functional magnetic resonance imaging
  • General linear model
  • Model assessment
  • Model building
  • Model selection
  • Quality assessment
  • Quality control
  • Temporal autocorrelation

ASJC Scopus subject areas

  • Clinical Neurology
  • Neuroscience(all)
  • Radiological and Ultrasound Technology

Cite this

Razavi, M., Grabowski, T. J., Vispoel, W. P., Monahan, P., Mehta, S., Eaton, B., & Bolinger, L. (2003). Model Assessment and Model Building in fMRI. Human Brain Mapping, 20(4), 227-238. https://doi.org/10.1002/hbm.10141

Model Assessment and Model Building in fMRI. / Razavi, Mehrdad; Grabowski, Thomas J.; Vispoel, Walter P.; Monahan, Patrick; Mehta, Sonya; Eaton, Brent; Bolinger, Lizann.

In: Human Brain Mapping, Vol. 20, No. 4, 12.2003, p. 227-238.

Research output: Contribution to journalArticle

Razavi, M, Grabowski, TJ, Vispoel, WP, Monahan, P, Mehta, S, Eaton, B & Bolinger, L 2003, 'Model Assessment and Model Building in fMRI', Human Brain Mapping, vol. 20, no. 4, pp. 227-238. https://doi.org/10.1002/hbm.10141
Razavi M, Grabowski TJ, Vispoel WP, Monahan P, Mehta S, Eaton B et al. Model Assessment and Model Building in fMRI. Human Brain Mapping. 2003 Dec;20(4):227-238. https://doi.org/10.1002/hbm.10141
Razavi, Mehrdad ; Grabowski, Thomas J. ; Vispoel, Walter P. ; Monahan, Patrick ; Mehta, Sonya ; Eaton, Brent ; Bolinger, Lizann. / Model Assessment and Model Building in fMRI. In: Human Brain Mapping. 2003 ; Vol. 20, No. 4. pp. 227-238.
@article{4cc8fff5f049445e85b78256f7f2b676,
title = "Model Assessment and Model Building in fMRI",
abstract = "Model quality is rarely assessed in fMRI data analyses and less often reported. This may have contributed to several shortcomings in the current fMRI data analyses, including: (1) Model mis-specification, leading to incorrect inference about the activation-maps, SPM{t} and SPM{F}; (2) Improper model selection based on the number of activated voxels, rather than on model quality; (3) Underutilization of systematic model building, resulting in the common but suboptimal practice of using only a single, pre-specified, usually over-simplified model; (4) Spatially homogenous modeling, neglecting the spatial heterogeneity of fMRI signal fluctuations; and (5) Lack of standards for formal model comparison, contributing to the high variability of fMRI results across studies and centers. To overcome these shortcomings, it is essential to assess and report the quality of the models used in the analysis. In this study, we applied images of the Durbin-Watson statistic (DW-map) and the coefficient of multiple determination (R2-map) as complementary tools to assess the validity as well as goodness of fit, i.e., quality, of models in fMRI data analysis. Higher quality models were built upon reduced models using classic model building. While inclusion of an appropriate variable in the model improved the quality of the model, inclusion of an inappropriate variable, i.e., model mis-specification, adversely affected it. Higher quality models, however, occasionally decreased the number of activated voxels, whereas lower quality or inappropriate models occasionally increased the number of activated voxels, indicating that the conventional approach to fMRI data analysis may yield sub-optimal or incorrect results. We propose that model quality maps become part of a broader package of maps for quality assessment in fMRI, facilitating validation, optimization, and standardization of fMRI result across studies and centers.",
keywords = "Functional magnetic resonance imaging, General linear model, Model assessment, Model building, Model selection, Quality assessment, Quality control, Temporal autocorrelation",
author = "Mehrdad Razavi and Grabowski, {Thomas J.} and Vispoel, {Walter P.} and Patrick Monahan and Sonya Mehta and Brent Eaton and Lizann Bolinger",
year = "2003",
month = "12",
doi = "10.1002/hbm.10141",
language = "English (US)",
volume = "20",
pages = "227--238",
journal = "Human Brain Mapping",
issn = "1065-9471",
publisher = "Wiley-Liss Inc.",
number = "4",

}

TY - JOUR

T1 - Model Assessment and Model Building in fMRI

AU - Razavi, Mehrdad

AU - Grabowski, Thomas J.

AU - Vispoel, Walter P.

AU - Monahan, Patrick

AU - Mehta, Sonya

AU - Eaton, Brent

AU - Bolinger, Lizann

PY - 2003/12

Y1 - 2003/12

N2 - Model quality is rarely assessed in fMRI data analyses and less often reported. This may have contributed to several shortcomings in the current fMRI data analyses, including: (1) Model mis-specification, leading to incorrect inference about the activation-maps, SPM{t} and SPM{F}; (2) Improper model selection based on the number of activated voxels, rather than on model quality; (3) Underutilization of systematic model building, resulting in the common but suboptimal practice of using only a single, pre-specified, usually over-simplified model; (4) Spatially homogenous modeling, neglecting the spatial heterogeneity of fMRI signal fluctuations; and (5) Lack of standards for formal model comparison, contributing to the high variability of fMRI results across studies and centers. To overcome these shortcomings, it is essential to assess and report the quality of the models used in the analysis. In this study, we applied images of the Durbin-Watson statistic (DW-map) and the coefficient of multiple determination (R2-map) as complementary tools to assess the validity as well as goodness of fit, i.e., quality, of models in fMRI data analysis. Higher quality models were built upon reduced models using classic model building. While inclusion of an appropriate variable in the model improved the quality of the model, inclusion of an inappropriate variable, i.e., model mis-specification, adversely affected it. Higher quality models, however, occasionally decreased the number of activated voxels, whereas lower quality or inappropriate models occasionally increased the number of activated voxels, indicating that the conventional approach to fMRI data analysis may yield sub-optimal or incorrect results. We propose that model quality maps become part of a broader package of maps for quality assessment in fMRI, facilitating validation, optimization, and standardization of fMRI result across studies and centers.

AB - Model quality is rarely assessed in fMRI data analyses and less often reported. This may have contributed to several shortcomings in the current fMRI data analyses, including: (1) Model mis-specification, leading to incorrect inference about the activation-maps, SPM{t} and SPM{F}; (2) Improper model selection based on the number of activated voxels, rather than on model quality; (3) Underutilization of systematic model building, resulting in the common but suboptimal practice of using only a single, pre-specified, usually over-simplified model; (4) Spatially homogenous modeling, neglecting the spatial heterogeneity of fMRI signal fluctuations; and (5) Lack of standards for formal model comparison, contributing to the high variability of fMRI results across studies and centers. To overcome these shortcomings, it is essential to assess and report the quality of the models used in the analysis. In this study, we applied images of the Durbin-Watson statistic (DW-map) and the coefficient of multiple determination (R2-map) as complementary tools to assess the validity as well as goodness of fit, i.e., quality, of models in fMRI data analysis. Higher quality models were built upon reduced models using classic model building. While inclusion of an appropriate variable in the model improved the quality of the model, inclusion of an inappropriate variable, i.e., model mis-specification, adversely affected it. Higher quality models, however, occasionally decreased the number of activated voxels, whereas lower quality or inappropriate models occasionally increased the number of activated voxels, indicating that the conventional approach to fMRI data analysis may yield sub-optimal or incorrect results. We propose that model quality maps become part of a broader package of maps for quality assessment in fMRI, facilitating validation, optimization, and standardization of fMRI result across studies and centers.

KW - Functional magnetic resonance imaging

KW - General linear model

KW - Model assessment

KW - Model building

KW - Model selection

KW - Quality assessment

KW - Quality control

KW - Temporal autocorrelation

UR - http://www.scopus.com/inward/record.url?scp=0344826551&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0344826551&partnerID=8YFLogxK

U2 - 10.1002/hbm.10141

DO - 10.1002/hbm.10141

M3 - Article

C2 - 14673806

AN - SCOPUS:0344826551

VL - 20

SP - 227

EP - 238

JO - Human Brain Mapping

JF - Human Brain Mapping

SN - 1065-9471

IS - 4

ER -