MARM: Multiscale adaptive regression models for neuroimaging data

Hongtu Zhu, Yimei Li, Joseph G. Ibrahim, Weili Lin, Dinggang Shen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We develop a novel statistical model, called multiscale adaptive regression model (MARM), for spatial and adaptive analysis of neuroimaging data. The primary motivation and application of the proposed methodology is statistical analysis of imaging data on the two-dimensional (2D) surface or in the 3D volume for various neuroimaging studies. The existing voxel-wise approach has several major limitations for the analyses of imaging data, underscoring the great need for methodological development. The voxel-wise approach essentially treats all voxels as independent units, whereas neuroimaging data are spatially correlated in nature and spatially contiguous regions of activation with rather sharp edges are usually expected. The initial smoothing step before the voxel-wise approach often blurs the image data near the edges of activated regions and thus it can dramatically increase the numbers of false positives and false negatives. The MARM, which is developed for addressing these limitations, has three key features in the analysis of imaging data: being spatial, being hierarchical, and being adaptive. The MARM builds a small sphere at each location (called voxel) and use these consecutively connected spheres across all voxels to capture spatial dependence among imaging observations. Then, the MARM builds hierarchically nested spheres by increasing the radius of a spherical neighborhood around each voxel and combine all the data in a given radius of each voxel with appropriate weights to adaptively calculate parameter estimates and test statistics. Theoretically, we first establish that the MARM outperforms classical voxel-wise approach. Simulation studies are used to demonstrate the methodology and examine the finite sample performance of the MARM. We apply our methods to the detection of spatial patterns of brain atrophy in a neuroimaging study of Alzheimer's disease. Our simulation studies with known ground truth confirm that the MARM significantly outperforms the voxel-wise methods.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages314-325
Number of pages12
Volume5636 LNCS
DOIs
Publication statusPublished - 2009 Sep 21
Externally publishedYes
Event21st International Conference on Information Processing in Medical Imaging, IPMI 2009 - Williamsburg, VA, United States
Duration: 2009 Jul 52009 Jul 10

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume5636 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other21st International Conference on Information Processing in Medical Imaging, IPMI 2009
CountryUnited States
CityWilliamsburg, VA
Period09/7/509/7/10

Fingerprint

Neuroimaging
Voxel
Regression Model
Imaging techniques
Imaging
Radius
Simulation Study
Spatial Dependence
Brain
Alzheimer's Disease
Statistical methods
Methodology
Spatial Pattern
Chemical activation
Spatial Data
Statistics
False Positive
Statistical Model
Test Statistic
Statistical Analysis

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Zhu, H., Li, Y., Ibrahim, J. G., Lin, W., & Shen, D. (2009). MARM: Multiscale adaptive regression models for neuroimaging data. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5636 LNCS, pp. 314-325). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5636 LNCS). https://doi.org/10.1007/978-3-642-02498-6_26

MARM : Multiscale adaptive regression models for neuroimaging data. / Zhu, Hongtu; Li, Yimei; Ibrahim, Joseph G.; Lin, Weili; Shen, Dinggang.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5636 LNCS 2009. p. 314-325 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5636 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Zhu, H, Li, Y, Ibrahim, JG, Lin, W & Shen, D 2009, MARM: Multiscale adaptive regression models for neuroimaging data. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 5636 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 5636 LNCS, pp. 314-325, 21st International Conference on Information Processing in Medical Imaging, IPMI 2009, Williamsburg, VA, United States, 09/7/5. https://doi.org/10.1007/978-3-642-02498-6_26
Zhu H, Li Y, Ibrahim JG, Lin W, Shen D. MARM: Multiscale adaptive regression models for neuroimaging data. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5636 LNCS. 2009. p. 314-325. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-02498-6_26
Zhu, Hongtu ; Li, Yimei ; Ibrahim, Joseph G. ; Lin, Weili ; Shen, Dinggang. / MARM : Multiscale adaptive regression models for neuroimaging data. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5636 LNCS 2009. pp. 314-325 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{053f886c42ee46aa8c70656c61e10333,
title = "MARM: Multiscale adaptive regression models for neuroimaging data",
abstract = "We develop a novel statistical model, called multiscale adaptive regression model (MARM), for spatial and adaptive analysis of neuroimaging data. The primary motivation and application of the proposed methodology is statistical analysis of imaging data on the two-dimensional (2D) surface or in the 3D volume for various neuroimaging studies. The existing voxel-wise approach has several major limitations for the analyses of imaging data, underscoring the great need for methodological development. The voxel-wise approach essentially treats all voxels as independent units, whereas neuroimaging data are spatially correlated in nature and spatially contiguous regions of activation with rather sharp edges are usually expected. The initial smoothing step before the voxel-wise approach often blurs the image data near the edges of activated regions and thus it can dramatically increase the numbers of false positives and false negatives. The MARM, which is developed for addressing these limitations, has three key features in the analysis of imaging data: being spatial, being hierarchical, and being adaptive. The MARM builds a small sphere at each location (called voxel) and use these consecutively connected spheres across all voxels to capture spatial dependence among imaging observations. Then, the MARM builds hierarchically nested spheres by increasing the radius of a spherical neighborhood around each voxel and combine all the data in a given radius of each voxel with appropriate weights to adaptively calculate parameter estimates and test statistics. Theoretically, we first establish that the MARM outperforms classical voxel-wise approach. Simulation studies are used to demonstrate the methodology and examine the finite sample performance of the MARM. We apply our methods to the detection of spatial patterns of brain atrophy in a neuroimaging study of Alzheimer's disease. Our simulation studies with known ground truth confirm that the MARM significantly outperforms the voxel-wise methods.",
author = "Hongtu Zhu and Yimei Li and Ibrahim, {Joseph G.} and Weili Lin and Dinggang Shen",
year = "2009",
month = "9",
day = "21",
doi = "10.1007/978-3-642-02498-6_26",
language = "English",
isbn = "3642024971",
volume = "5636 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "314--325",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - MARM

T2 - Multiscale adaptive regression models for neuroimaging data

AU - Zhu, Hongtu

AU - Li, Yimei

AU - Ibrahim, Joseph G.

AU - Lin, Weili

AU - Shen, Dinggang

PY - 2009/9/21

Y1 - 2009/9/21

N2 - We develop a novel statistical model, called multiscale adaptive regression model (MARM), for spatial and adaptive analysis of neuroimaging data. The primary motivation and application of the proposed methodology is statistical analysis of imaging data on the two-dimensional (2D) surface or in the 3D volume for various neuroimaging studies. The existing voxel-wise approach has several major limitations for the analyses of imaging data, underscoring the great need for methodological development. The voxel-wise approach essentially treats all voxels as independent units, whereas neuroimaging data are spatially correlated in nature and spatially contiguous regions of activation with rather sharp edges are usually expected. The initial smoothing step before the voxel-wise approach often blurs the image data near the edges of activated regions and thus it can dramatically increase the numbers of false positives and false negatives. The MARM, which is developed for addressing these limitations, has three key features in the analysis of imaging data: being spatial, being hierarchical, and being adaptive. The MARM builds a small sphere at each location (called voxel) and use these consecutively connected spheres across all voxels to capture spatial dependence among imaging observations. Then, the MARM builds hierarchically nested spheres by increasing the radius of a spherical neighborhood around each voxel and combine all the data in a given radius of each voxel with appropriate weights to adaptively calculate parameter estimates and test statistics. Theoretically, we first establish that the MARM outperforms classical voxel-wise approach. Simulation studies are used to demonstrate the methodology and examine the finite sample performance of the MARM. We apply our methods to the detection of spatial patterns of brain atrophy in a neuroimaging study of Alzheimer's disease. Our simulation studies with known ground truth confirm that the MARM significantly outperforms the voxel-wise methods.

AB - We develop a novel statistical model, called multiscale adaptive regression model (MARM), for spatial and adaptive analysis of neuroimaging data. The primary motivation and application of the proposed methodology is statistical analysis of imaging data on the two-dimensional (2D) surface or in the 3D volume for various neuroimaging studies. The existing voxel-wise approach has several major limitations for the analyses of imaging data, underscoring the great need for methodological development. The voxel-wise approach essentially treats all voxels as independent units, whereas neuroimaging data are spatially correlated in nature and spatially contiguous regions of activation with rather sharp edges are usually expected. The initial smoothing step before the voxel-wise approach often blurs the image data near the edges of activated regions and thus it can dramatically increase the numbers of false positives and false negatives. The MARM, which is developed for addressing these limitations, has three key features in the analysis of imaging data: being spatial, being hierarchical, and being adaptive. The MARM builds a small sphere at each location (called voxel) and use these consecutively connected spheres across all voxels to capture spatial dependence among imaging observations. Then, the MARM builds hierarchically nested spheres by increasing the radius of a spherical neighborhood around each voxel and combine all the data in a given radius of each voxel with appropriate weights to adaptively calculate parameter estimates and test statistics. Theoretically, we first establish that the MARM outperforms classical voxel-wise approach. Simulation studies are used to demonstrate the methodology and examine the finite sample performance of the MARM. We apply our methods to the detection of spatial patterns of brain atrophy in a neuroimaging study of Alzheimer's disease. Our simulation studies with known ground truth confirm that the MARM significantly outperforms the voxel-wise methods.

UR - http://www.scopus.com/inward/record.url?scp=70349094537&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70349094537&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-02498-6_26

DO - 10.1007/978-3-642-02498-6_26

M3 - Conference contribution

SN - 3642024971

SN - 9783642024979

VL - 5636 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 314

EP - 325

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -