Sparse clustering with resampling for subject classification in PET amyloid imaging studies

Wenzhu Bi Mowrey, George C. Tseng, Lisa A. Weissfeld, Julie C. Price

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

2 Citations (Scopus)

Abstract

Sparse k-means clustering (Sparse-kM) can exclude uninformative variables and yield reliable parsimonious clustering results, especially for p≫n. In this work, Sparse-kM and data resampling were combined to identify variables of greatest interest and define confidence levels for the clustering. The method was evaluated by statistical simulation and applied to PiB PET amyloid imaging data to identify normal control (NC) subjects with (+) or without (-) evidence of amyloid, i.e., PiB(+/-). Simulations. A dataset of n=60 observations (3 groups of 20) and p=500 variables was generated for each simulation run; only 50 variables were truly different across groups. The dataset was resampled 20 times, Sparse-kM was applied to each sample and average variable weights were calculated. Probabilities of cluster membership, also called confidence levels, were computed (n=60). Simulations were performed 250 times. The 50 truly different variables were identified by variable weights that were 13-32 times greater than those for the 450 uninformative variables. Human Data. For the PiB PET dataset, images (ECAT HR+, 10-15 mCi, 90 min) were acquired for 64 cognitively normal subjects (74.1±5.4 yrs). Parametric PiB distribution volume ratio images were generated (Logan method, cerebellum reference) and normalized to the MNI template (SPM8) to produce a dataset of n=64 subjects and p=343,099 voxels/image. The dataset was resampled 10 times and Sparse-kM was applied. An average voxel weight image was computed that indicated cortical areas of greatest interest that included precuneus and frontal cortex; these are key areas linked to early amyloid deposition. Seven of 64 subjects were identified as PiB(+) and 47 as PiB(-) with confidence = 90%, where another subject was PiB(+) at lower confidence (80%) and the other 9 subjects were PiB(-) at confidence in the range of 50-70%. In conclusion, Sparse-kM with resampling can help to establish confidence levels for clustering when p≫n and may be a promising method for revealing informative voxels/spatial patterns that distinguish levels of amyloid load, including that at the transitional amyloid +/- boundary.

Original languageEnglish (US)
Title of host publicationIEEE Nuclear Science Symposium Conference Record
Pages3108-3111
Number of pages4
DOIs
StatePublished - 2012
Externally publishedYes
Event2011 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2011 - Valencia, Spain
Duration: Oct 23 2011Oct 29 2011

Other

Other2011 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2011
CountrySpain
CityValencia
Period10/23/1110/29/11

Fingerprint

Amyloid
Cluster Analysis
confidence
Weights and Measures
simulation
cerebellum
Parietal Lobe
cortexes
Frontal Lobe
Cerebellum
templates
Datasets

ASJC Scopus subject areas

  • Radiation
  • Nuclear and High Energy Physics
  • Radiology Nuclear Medicine and imaging

Cite this

Mowrey, W. B., Tseng, G. C., Weissfeld, L. A., & Price, J. C. (2012). Sparse clustering with resampling for subject classification in PET amyloid imaging studies. In IEEE Nuclear Science Symposium Conference Record (pp. 3108-3111). [6152564] https://doi.org/10.1109/NSSMIC.2011.6152564

Sparse clustering with resampling for subject classification in PET amyloid imaging studies. / Mowrey, Wenzhu Bi; Tseng, George C.; Weissfeld, Lisa A.; Price, Julie C.

IEEE Nuclear Science Symposium Conference Record. 2012. p. 3108-3111 6152564.

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

Mowrey, WB, Tseng, GC, Weissfeld, LA & Price, JC 2012, Sparse clustering with resampling for subject classification in PET amyloid imaging studies. in IEEE Nuclear Science Symposium Conference Record., 6152564, pp. 3108-3111, 2011 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2011, Valencia, Spain, 10/23/11. https://doi.org/10.1109/NSSMIC.2011.6152564
Mowrey WB, Tseng GC, Weissfeld LA, Price JC. Sparse clustering with resampling for subject classification in PET amyloid imaging studies. In IEEE Nuclear Science Symposium Conference Record. 2012. p. 3108-3111. 6152564 https://doi.org/10.1109/NSSMIC.2011.6152564
Mowrey, Wenzhu Bi ; Tseng, George C. ; Weissfeld, Lisa A. ; Price, Julie C. / Sparse clustering with resampling for subject classification in PET amyloid imaging studies. IEEE Nuclear Science Symposium Conference Record. 2012. pp. 3108-3111
@inproceedings{0d522c97264e43918945568a5aa94ecc,
title = "Sparse clustering with resampling for subject classification in PET amyloid imaging studies",
abstract = "Sparse k-means clustering (Sparse-kM) can exclude uninformative variables and yield reliable parsimonious clustering results, especially for p≫n. In this work, Sparse-kM and data resampling were combined to identify variables of greatest interest and define confidence levels for the clustering. The method was evaluated by statistical simulation and applied to PiB PET amyloid imaging data to identify normal control (NC) subjects with (+) or without (-) evidence of amyloid, i.e., PiB(+/-). Simulations. A dataset of n=60 observations (3 groups of 20) and p=500 variables was generated for each simulation run; only 50 variables were truly different across groups. The dataset was resampled 20 times, Sparse-kM was applied to each sample and average variable weights were calculated. Probabilities of cluster membership, also called confidence levels, were computed (n=60). Simulations were performed 250 times. The 50 truly different variables were identified by variable weights that were 13-32 times greater than those for the 450 uninformative variables. Human Data. For the PiB PET dataset, images (ECAT HR+, 10-15 mCi, 90 min) were acquired for 64 cognitively normal subjects (74.1±5.4 yrs). Parametric PiB distribution volume ratio images were generated (Logan method, cerebellum reference) and normalized to the MNI template (SPM8) to produce a dataset of n=64 subjects and p=343,099 voxels/image. The dataset was resampled 10 times and Sparse-kM was applied. An average voxel weight image was computed that indicated cortical areas of greatest interest that included precuneus and frontal cortex; these are key areas linked to early amyloid deposition. Seven of 64 subjects were identified as PiB(+) and 47 as PiB(-) with confidence = 90{\%}, where another subject was PiB(+) at lower confidence (80{\%}) and the other 9 subjects were PiB(-) at confidence in the range of 50-70{\%}. In conclusion, Sparse-kM with resampling can help to establish confidence levels for clustering when p≫n and may be a promising method for revealing informative voxels/spatial patterns that distinguish levels of amyloid load, including that at the transitional amyloid +/- boundary.",
author = "Mowrey, {Wenzhu Bi} and Tseng, {George C.} and Weissfeld, {Lisa A.} and Price, {Julie C.}",
year = "2012",
doi = "10.1109/NSSMIC.2011.6152564",
language = "English (US)",
isbn = "9781467301183",
pages = "3108--3111",
booktitle = "IEEE Nuclear Science Symposium Conference Record",

}

TY - GEN

T1 - Sparse clustering with resampling for subject classification in PET amyloid imaging studies

AU - Mowrey, Wenzhu Bi

AU - Tseng, George C.

AU - Weissfeld, Lisa A.

AU - Price, Julie C.

PY - 2012

Y1 - 2012

N2 - Sparse k-means clustering (Sparse-kM) can exclude uninformative variables and yield reliable parsimonious clustering results, especially for p≫n. In this work, Sparse-kM and data resampling were combined to identify variables of greatest interest and define confidence levels for the clustering. The method was evaluated by statistical simulation and applied to PiB PET amyloid imaging data to identify normal control (NC) subjects with (+) or without (-) evidence of amyloid, i.e., PiB(+/-). Simulations. A dataset of n=60 observations (3 groups of 20) and p=500 variables was generated for each simulation run; only 50 variables were truly different across groups. The dataset was resampled 20 times, Sparse-kM was applied to each sample and average variable weights were calculated. Probabilities of cluster membership, also called confidence levels, were computed (n=60). Simulations were performed 250 times. The 50 truly different variables were identified by variable weights that were 13-32 times greater than those for the 450 uninformative variables. Human Data. For the PiB PET dataset, images (ECAT HR+, 10-15 mCi, 90 min) were acquired for 64 cognitively normal subjects (74.1±5.4 yrs). Parametric PiB distribution volume ratio images were generated (Logan method, cerebellum reference) and normalized to the MNI template (SPM8) to produce a dataset of n=64 subjects and p=343,099 voxels/image. The dataset was resampled 10 times and Sparse-kM was applied. An average voxel weight image was computed that indicated cortical areas of greatest interest that included precuneus and frontal cortex; these are key areas linked to early amyloid deposition. Seven of 64 subjects were identified as PiB(+) and 47 as PiB(-) with confidence = 90%, where another subject was PiB(+) at lower confidence (80%) and the other 9 subjects were PiB(-) at confidence in the range of 50-70%. In conclusion, Sparse-kM with resampling can help to establish confidence levels for clustering when p≫n and may be a promising method for revealing informative voxels/spatial patterns that distinguish levels of amyloid load, including that at the transitional amyloid +/- boundary.

AB - Sparse k-means clustering (Sparse-kM) can exclude uninformative variables and yield reliable parsimonious clustering results, especially for p≫n. In this work, Sparse-kM and data resampling were combined to identify variables of greatest interest and define confidence levels for the clustering. The method was evaluated by statistical simulation and applied to PiB PET amyloid imaging data to identify normal control (NC) subjects with (+) or without (-) evidence of amyloid, i.e., PiB(+/-). Simulations. A dataset of n=60 observations (3 groups of 20) and p=500 variables was generated for each simulation run; only 50 variables were truly different across groups. The dataset was resampled 20 times, Sparse-kM was applied to each sample and average variable weights were calculated. Probabilities of cluster membership, also called confidence levels, were computed (n=60). Simulations were performed 250 times. The 50 truly different variables were identified by variable weights that were 13-32 times greater than those for the 450 uninformative variables. Human Data. For the PiB PET dataset, images (ECAT HR+, 10-15 mCi, 90 min) were acquired for 64 cognitively normal subjects (74.1±5.4 yrs). Parametric PiB distribution volume ratio images were generated (Logan method, cerebellum reference) and normalized to the MNI template (SPM8) to produce a dataset of n=64 subjects and p=343,099 voxels/image. The dataset was resampled 10 times and Sparse-kM was applied. An average voxel weight image was computed that indicated cortical areas of greatest interest that included precuneus and frontal cortex; these are key areas linked to early amyloid deposition. Seven of 64 subjects were identified as PiB(+) and 47 as PiB(-) with confidence = 90%, where another subject was PiB(+) at lower confidence (80%) and the other 9 subjects were PiB(-) at confidence in the range of 50-70%. In conclusion, Sparse-kM with resampling can help to establish confidence levels for clustering when p≫n and may be a promising method for revealing informative voxels/spatial patterns that distinguish levels of amyloid load, including that at the transitional amyloid +/- boundary.

UR - http://www.scopus.com/inward/record.url?scp=84858651751&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84858651751&partnerID=8YFLogxK

U2 - 10.1109/NSSMIC.2011.6152564

DO - 10.1109/NSSMIC.2011.6152564

M3 - Conference contribution

AN - SCOPUS:84858651751

SN - 9781467301183

SP - 3108

EP - 3111

BT - IEEE Nuclear Science Symposium Conference Record

ER -