Parcellationofthethalamusbasedongraymattervolumecovarianceinhealthymales
10.3969/j.issn.1002-1671.2019.03.002
- VernacularTitle:基于灰质体积共变模式的正常男性丘脑分割
- Author:
Ying LI
1
;
Yang JI
;
Xueli WANG
Author Information
1. 西安医学院第一附属医院影像科
- Keywords:
graymattervolumecovariance;
KGmeansclustering;
crossvalidate;
thalamus;
parcellation
- From:
Journal of Practical Radiology
2019;35(3):345-348
- CountryChina
- Language:Chinese
-
Abstract:
Objective Toexplorethefeasibilityandaccuracyofsegmentationofthethalamusinhealthymaleswithgraymatter volumecovariance(GMVC)covariancepattern.Methods T1 3DstructuralMRIdatawerecollectedin103 maleparticipants [age from19to50y,mean＝(35.5±7.7)y,educationfrom6to16y,mean＝(10.5±2.3)y].AfterpreprocessedbyCAT12softwarewith standardprocedure (registration,segmentation,modulationandsmooth),thevoxelGwisebilateralthalamusgray mattervolume (GMV)andthewholebrain’svoxelGwiseGMVdatawereextractedtocalculatethecovariancematrixofallparticipants.ExtendedKGmeans clusteringalgorithm wasappliedtosegmentthalamusintosubGareas.RandomsubGsamplingandcrossingGvalidationtechniquewere usedtogenerate100subGdatasetsandverifythesimilarityindex.Cramers’V (CV)wascalculatedandtheKrelevanttothehighest meanofCV wasdeterminedastheoptimalnumberofsubGareasforthalamus.Results Theleftandrightthalamuswereoptimally segmentedinto11and13subGareasbyGMVcovariancepattern.BoththeleftandrightthalamuscomprisesAnteriorNuclear,ventral anteriornucleus,lateraldorsalnucleus,lateralposteriornucleus,ventrallateralnucleus,ventralposteriormedialnucleus,ventralposterior lateralnucleus,midlinenucleus,medialdorsalnucleusandpulvinar.Additionally,fortherightthalamus,anteriornucleargroupwas dividedinto2parts,dorsalpartandventralpart,andthemidlinenucleusintomedialpartandlateralpart.Thelateraldorsalandlateral posteriornucleusweremergedintolateralnucleargroup.Fortheleftthalamus,thelateralposteriornucleuswasfurtherdividedinto2 parts,anteriorpartandposteriorpart.Thesegmentationresultswerealmostalinewiththeanatomical14partitionofthalamus.Conclusion TheparcellationbasedonGMVCcovariancepatternisaneffectiveandreliabletoolforsegmentationofbraingraymatternucleus. Ourfindingsprovideexperimentbasisforthalamusfunctionalandstructuralstudyinthefutureandoptimaltreatmenttargetforneurosurgery anddeepbrainstimulustherapy.
