Classification of multiple sclerosis and non-specific white matter lesions using spherical harmonics descriptors

Yeqi Wang; Madison Hansen; Darin Okuda; Andrew Wilson; Xiaohu Guo

doi:10.1145/3191801.3191806

Classification of multiple sclerosis and non-specific white matter lesions using spherical harmonics descriptors

Yeqi Wang, Madison Hansen, Darin Okuda, Andrew Wilson, Xiaohu Guo

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Scopus citations

Abstract

Multiple Sclerosis (MS) and Non-Specific White Matter (NSWM) lesion classification is a traditional problem in neurology. In this paper, we propose a machine learning method to predict MS/NSWM lesion types based on segmented 3D lesion models from clinical MRI images. A spherical harmonics shape descriptor is used to express lesion shape as feature vectors. We generate a parametrization mapping from a 3D lesion to a sphere and then extract spherical harmonics features as descriptor based on that mapping. This descriptor conveys shape difference properties of MS/NSWM lesion which can be trained to predict unknown lesions using machine learning models such as boosting trees, support vector machines (SVMs), logistic regression, and so on. Experiments demonstrate that our 3D model feature representation enables significant performance on MS/NSWM lesions in their classification.

Original language	English (US)
Title of host publication	Proceedings of the 3rd International Workshop on Interactive and Spatial Computing, IWISC 2018
Publisher	Association for Computing Machinery
Pages	97-102
Number of pages	6
ISBN (Electronic)	9781450354394
DOIs	https://doi.org/10.1145/3191801.3191806
State	Published - Apr 12 2018
Event	3rd International Workshop on Interactive and Spatial Computing, IWISC 2018 - Richardson, United States Duration: Apr 12 2018 → Apr 13 2018

Other

Other	3rd International Workshop on Interactive and Spatial Computing, IWISC 2018
Country/Territory	United States
City	Richardson
Period	4/12/18 → 4/13/18

Keywords

3D shape analysis
Machine learning
Multiple sclerosis
Non-specific white matter
Spherical harmonics

ASJC Scopus subject areas

Human-Computer Interaction
Computer Networks and Communications
Computer Vision and Pattern Recognition
Software

Access to Document

10.1145/3191801.3191806

Cite this

Classification of multiple sclerosis and non-specific white matter lesions using spherical harmonics descriptors. / Wang, Yeqi; Hansen, Madison; Okuda, Darin et al.
Proceedings of the 3rd International Workshop on Interactive and Spatial Computing, IWISC 2018. Association for Computing Machinery, 2018. p. 97-102.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Wang, Y, Hansen, M, Okuda, D, Wilson, A & Guo, X 2018, Classification of multiple sclerosis and non-specific white matter lesions using spherical harmonics descriptors. in Proceedings of the 3rd International Workshop on Interactive and Spatial Computing, IWISC 2018. Association for Computing Machinery, pp. 97-102, 3rd International Workshop on Interactive and Spatial Computing, IWISC 2018, Richardson, United States, 4/12/18. https://doi.org/10.1145/3191801.3191806

@inproceedings{9ea013c8441a424bafd1c385ada05250,

title = "Classification of multiple sclerosis and non-specific white matter lesions using spherical harmonics descriptors",

abstract = "Multiple Sclerosis (MS) and Non-Specific White Matter (NSWM) lesion classification is a traditional problem in neurology. In this paper, we propose a machine learning method to predict MS/NSWM lesion types based on segmented 3D lesion models from clinical MRI images. A spherical harmonics shape descriptor is used to express lesion shape as feature vectors. We generate a parametrization mapping from a 3D lesion to a sphere and then extract spherical harmonics features as descriptor based on that mapping. This descriptor conveys shape difference properties of MS/NSWM lesion which can be trained to predict unknown lesions using machine learning models such as boosting trees, support vector machines (SVMs), logistic regression, and so on. Experiments demonstrate that our 3D model feature representation enables significant performance on MS/NSWM lesions in their classification.",

keywords = "3D shape analysis, Machine learning, Multiple sclerosis, Non-specific white matter, Spherical harmonics",

author = "Yeqi Wang and Madison Hansen and Darin Okuda and Andrew Wilson and Xiaohu Guo",

year = "2018",

month = apr,

day = "12",

doi = "10.1145/3191801.3191806",

language = "English (US)",

pages = "97--102",

booktitle = "Proceedings of the 3rd International Workshop on Interactive and Spatial Computing, IWISC 2018",

publisher = "Association for Computing Machinery",

note = "3rd International Workshop on Interactive and Spatial Computing, IWISC 2018 ; Conference date: 12-04-2018 Through 13-04-2018",

}

TY - GEN

T1 - Classification of multiple sclerosis and non-specific white matter lesions using spherical harmonics descriptors

AU - Wang, Yeqi

AU - Hansen, Madison

AU - Okuda, Darin

AU - Wilson, Andrew

AU - Guo, Xiaohu

PY - 2018/4/12

Y1 - 2018/4/12

N2 - Multiple Sclerosis (MS) and Non-Specific White Matter (NSWM) lesion classification is a traditional problem in neurology. In this paper, we propose a machine learning method to predict MS/NSWM lesion types based on segmented 3D lesion models from clinical MRI images. A spherical harmonics shape descriptor is used to express lesion shape as feature vectors. We generate a parametrization mapping from a 3D lesion to a sphere and then extract spherical harmonics features as descriptor based on that mapping. This descriptor conveys shape difference properties of MS/NSWM lesion which can be trained to predict unknown lesions using machine learning models such as boosting trees, support vector machines (SVMs), logistic regression, and so on. Experiments demonstrate that our 3D model feature representation enables significant performance on MS/NSWM lesions in their classification.

AB - Multiple Sclerosis (MS) and Non-Specific White Matter (NSWM) lesion classification is a traditional problem in neurology. In this paper, we propose a machine learning method to predict MS/NSWM lesion types based on segmented 3D lesion models from clinical MRI images. A spherical harmonics shape descriptor is used to express lesion shape as feature vectors. We generate a parametrization mapping from a 3D lesion to a sphere and then extract spherical harmonics features as descriptor based on that mapping. This descriptor conveys shape difference properties of MS/NSWM lesion which can be trained to predict unknown lesions using machine learning models such as boosting trees, support vector machines (SVMs), logistic regression, and so on. Experiments demonstrate that our 3D model feature representation enables significant performance on MS/NSWM lesions in their classification.

KW - 3D shape analysis

KW - Machine learning

KW - Multiple sclerosis

KW - Non-specific white matter

KW - Spherical harmonics

UR - http://www.scopus.com/inward/record.url?scp=85048830961&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85048830961&partnerID=8YFLogxK

U2 - 10.1145/3191801.3191806

DO - 10.1145/3191801.3191806

M3 - Conference contribution

AN - SCOPUS:85048830961

SP - 97

EP - 102

BT - Proceedings of the 3rd International Workshop on Interactive and Spatial Computing, IWISC 2018

PB - Association for Computing Machinery

T2 - 3rd International Workshop on Interactive and Spatial Computing, IWISC 2018

Y2 - 12 April 2018 through 13 April 2018

ER -

Classification of multiple sclerosis and non-specific white matter lesions using spherical harmonics descriptors

Abstract

Other

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this