www.ksmrm.org JKSMRM 16(3) : 243-252, 2012 Print ISSN 1226-9751 Original Article 다기관의자기공명영상장치에서측정된뇌조직체적의재현성분석 정원범 1 강민재 1 손두범 2 김영주 2 이영민 3 김영훈 4 은충기 5 문치웅 1 1 인제대학교의용공학과, UHRC 2 해운대백병원영상의학과 3 부산대학교의과대학부산대학교병원정신과 4 인제대학교의과대학해운대백병원정신과 5 인제대학교의과대학해운대백병원영상의학과 목적 : 뇌자기공명영상기반의다기관연구구축을위해장치차이에따른동일인의자기공명영상에서측정한뇌조직의체적차이를확인함으로써체적변화에따른뇌질환의임상적진단에대한유의성평가자료를마련하고자하였다. 대상및방법 : 정상남녀 10명에대해시스템사양이같은장치 2대를포함한전체 4대의자기공명장치에서 3D T1 강조영상을획득하여, 자동화분할 S/W를통해회백질및하부구조를분할하였으며, 급내상관계수와변이계수를통해체적의일치성및차이를확인하고, 통계적유효성을확인하기위한시스템간체적차오프셋을측정하였다. 결과 : 모든영역에대해 MRI 장치간평균 0.837의체적일치성과평균 4.310% 의체적변화범위가측정되었고, 같은 MRI 장치간전체영역에서평균 3.611% 의체적차이가발생하였으며, 평균 7.079% 의오프셋이요구되었다. 또한시스템사양및영상획득인자가다른 MRI 장치간전체영역에대해약 5.785% 의체적차이가발생하였으며, 이에대한오프셋은약 11.340% 로확인되었다. 결론 : 다기관뇌연구의기반을구축하기위해서는 MRI 장치간에발생되는체적측정결과의변동을최소화하는하드웨어와소프트웨어의최적화가필요하며, 데이터관리및고속영상처리방법또한필요한것으로판단된다. 서 론 복잡한신경망으로구성된인간의뇌를비침습적으로관찰하고진단할수있는자기공명영상 (Magnetic Resonance Imaging, MRI) 은구조적또는기능적영상기법을이용하여병리학적변화를평가할수있다 (1-4). 노화에따른전반적인뇌의위축과비교하여 (5-7), 알츠하이머성치매 (Alzheimer s disease) (8), 파킨슨병 (Parkinson s disease) (9), 헌팅턴병 (Huntington s disease) (10) 등의신경퇴행성뇌질환및정신분열증 (Schizophrenia) (11), 우울증 (Depressive disorder) (12), 조울증 (Bipolar disorder) (13) 과같은정신질환에 Received; August 13, 2012 Revised; September 24, 2012 Accepted; December 13, 2012 Corresponding author : Chi Woong Mun, Ph.D., Department of Biomedical Engineering and UHRC, Inje University, 607, Obang-dong, Gimhae-si, Gyeongnam 621-749, South Korea. Tel. 82-55-320-3297, Fax. 82-55-327-3292 E-mail : mcw@inje.ac.kr 의해뇌구조의특정영역에서의변화와함께전체적인뇌의위축이발생하는것으로알려져있다 (8-14). 또한 MRI 를이용한구조적영상분석의경우, 알츠하이머성치매에대해뇌의구조적연결성을평가하는확산텐서영상 (Diffusion Tensor Imaging, DTI) 및기능적관계를나타내는기능적자기공명영상 (Functional MRI, fmri) 보다질환의진행과정에따른뇌의병리학적변화 (intermediate phenotypes) 를특징적으로나타내는데유용한것으로알려져있다 (15). 이에따라임상적증상 (Clinical symptom) 이감지되기이전의조기진단을통한질환진행의지연및치료계획수립을위해연부조직에대한대조도가뛰어난 MR 영상을이용하여뇌구조의변화양상을분석하고, 뇌질환의특성을직관적으로나타낼수있는영상마커 (image-marker) 의정립이필요하다. 하지만현재뇌질환에대한연구는영상학적및임상적관점에따른분석방법의차이로인해같은분야의연구임에도상이한결과를도출할수있고 (16-18), 분석대상의수가적거나단발성연구 (Cross-sectional study) 를수행함에따라그결과의신뢰성에한계를가질수있다 (19, 20). 그러므로통계적검증력을확보하기위해통일된연구 243
JKSMRM 16(3) : 243-252, 2012 프로토콜및대량의데이터를이용한영상분석이요구되며, 동일한연구목적을가진여러기관이참여하는공동연구 (multi-center study) 가제안된다 (21, 22). 다기관연구는단일기관연구에비해대상자를수집하는데있어시간적으로효율적일뿐만아니라, 특정지역내의대상자만을선별하는것이아니므로질환의일반성을가질수있다 (23, 24). 그러나영상의질은 MRI 제조사, 주자장강도등의시스템종류및영상획득조건에따라다르므로질환의영향으로인한실질적변화추정에대해혼동을야기할수있고 (25, 26), 그에따라통계적인결론을도출하는데제한을줄수있다 (27). 따라서시스템에의한측정결과차이를최소화하고정량적인영상분석을도출하기위해서는반드시영상획득인자, MRI 장치의사양및특성에따른영향이확인되어야한다. 본연구에서는다기관연구구축을위한기초연구로서자동화분할 S/W 를이용하여동일인에대해다른시스템의자기공명영상장치및영상획득인자에서발생하는각영역별체적의차이를확인하고, 이를기반으로질환에따른변화를감지할수있는시스템간의유의성을평가하고자하였다. 대상및방법 가. 영상획득정상성인남녀 10 명을 ( 남 / 녀 =6/4 명, 나이 =26.60±3.86 세 ) 대상으로다른기관에설치되어있는 4 대의임상용 MRI 장치에서 3 차원 T1 강조영상 (3D T1 weighted imaging) 을획득하였다. 본연구에사용된 MRI 장치는다음과같으며, 동일인에대해획득된영상은기술된 MRI 장치의순서대로 MR(a), MR(b), MR(c), MR(d) 라정의하였다 : 2 대의 Philips Achieva Tx 3T (Philips, Healthcare, Best, The Netherlands), Siemens Verio 3T (Siemens, Erlangen, Germany), GE Signa HDxt 1.5T (GE Healthcare, Waukesha, WI, USA). 영상의획득조건은 Table 1 에나타내었다. 나. 영상분석획득된 MR 영상은자동화분할 S/W 를사용하여 Figure 1 과같은영역의분할및체적분석을실시하였다. 전반적인뇌의변화를확인하기위해회백질 (gray matter, GM) 영역과퇴행성뇌질환의영향으로인해유의한변화가많은하부구조를관심영역으로선정하였으며 (8-13), 이러한영역들에서동일인에대해 MRI 장치의시스템적차이로인한체적변화를확인하였다. SPM 8 (Statistical Parametric Mapping 8, Welcome Department of Neuroimaging, London, UK, http://www.fil.ion.ucl.ac.uk/spm)(28) 에서연동가능한 VBM 8 (http://dbm.neuro.unijena.de) (29) S/W 를이용하여 WM, GM 및뇌척수액 (cerebrospinal fluid, CSF) 등조직별영역분할및체적분석을하였고, 기저핵 (basal ganglia) 및해마 (hippocampus) 등의뇌하부구조에대한분석은 FreeSurfer v.5.1 (MGH, Harvard, Boston, MA, USA, http://surfer.nmr.mgh.harvard.edu) (30, 31) 를이용하여수행하였다. SPM 및 FreeSurfer 는자동적영상분할 S/W 로서뇌조직별영역의체적크기가정해진가상 MR 영상 (simulated brain MR database) (32) 을이용한영상분할결과와실제 MR 영상을이용한수동적관심영역분할방법과자동적관심영역분할방법에대한비교결과를미루어보아, SPM 은하부구조에대한영역분할은할수없지만뇌조직별영역에대한분할정확성이우수하며 (33), 하부구조에대한분할및체적분석은 FreeSurfer 가우수한것으로알려져있다 (34, 35). SPM/VBM8 을이용한 MR 영상처리는뇌조직에따른영상신호강도의차이에대한영역별신호분포조절, 신호잡음제거, 불균일한영상밝기조절, 뇌표준판 (IXI-brain template, http://dbm.neuro.uni-jena.de) 으로의공간정규화및영역별조직분류화등의영상분할을위한조건을기본값 Table 1. MR Scanners and Protocols MRI Scanners Philips Achieva Tx Siemens Verio GE Signa HDxt Field Strength (T) 3 3 1.5 Sequence Type 1) 3D T1-TFE 3D T1-MPRAGE 3D T1-SPGR TR (ms)/te (ms) 8.6/3.96 1800/2.22 5.7/1.69 TI (ms) 1300 1000 - Flip Angle( ) 8 9 10 FOV (mm 2 ) 256 256 256 256 250 250 Matrix 256 256 256 256 256 256 Voxel size (mm 3 ) 1 1 1 1 1 1 0.98 0.98 1 Slice thickness (mm) 1 1 1 Angulation Sagittal Sagittal Sagittal Total scan time (min:s) 6:18 7:42 7:13 Note: 1) TFE = turbo field echo, MPRAGE = magnetization-prepared rapid acquisition gradient echo, SPGR = spoiled gradient echo 244
다기관의자기공명영상장치에서측정된뇌조직체적의재현성분석 정원범외 으로설정하여 (Default parameter settings) 일률적으로실시하였다. 또한, 뇌하부구조에대한자동적영상분할은 FreeSurfer 의 Cross-sectional stream 과정을이용하였으며, SPM 과같이분할이전의전처리과정을실시한후, Talairach 좌표공간으로공간정규화된뇌영상을수동분할된하부구조표준판 (Manual labeling by Center for Morphometric Analysis) 과의공간적관계에대한정보와영상의신호강도의확률적분포정보를결합하여반복적계산방법 (Iterative method) 을통해영역분할을수행하였다. 분할된영역의체적은공간정규화과정의변환행렬을역변환하여원영상의공간위치 (Native image space) 에서측정하였다. 다. 통계분석동일대상자의뇌조직및하부구조분할에의해측정된뇌조직의체적에대해 4 개의 MRI 장치간 (inter MR scanner) 일치도및그에따른체적결과의신뢰성을평가하기위해급내상관계수 (Intra-class Correlation Coefficient, ICC) 를이용하였다. ICC 측정모형은혼합이 원배치 (two-way mixed model) 로설정하고, 유형은일치 (consistency) 로구성하여, 4 개의 MRI 장치는고정된변수 (fixed factor) 로지정하고대상자는확률적으로 (randomly) 선택되는변수로적용하여신뢰도를계산하였다. 또한, 각대상자들의영역별체적에대해변이계수 (Coefficient of variation, CV) 를측정하여 4 개의 MRI 장치에서발생되는전반적인체적의차이를확인하였으며, 통계적으로유의한차이가있는지판단하기위해 Bonferoni 사후검증을통한반복측정분산분석 (repeated measures ANOVA) 을수행하였다. 추가적으로 Eq. [1] 과같이정의된 PVD (Percent volume difference) 를이용하여각 MRI 장치간에발생하는체적의차이를측정하였고, 이를기초로뇌조직의체적변화를측정하기위해동일대상자에서최소 2 개의다른 MRI 장치에서요구되는각영역에서의체적의차이에오프셋 (p < 0.05) 의범위를측정하였다. MR 영상을이용하여측정된관심뇌조직들의체적 (V 1 ) 이실제의각체적에대해정규분포를형성한다고가정할때, 동일인에대해 2 개의 MRI 장치에서획득된관심영역의체적 V 1, V 2 의분산이각각 V 1 과 V 2 이면그차이 Fig. 1. Examples of brain segmentation for region of interest: (a) gray matter (b) caudate nucleus (red), putamen (pink) and thalamus (green) (c) hippocampus (yellow) and amygdala (sky blue) (d) lateral ventricle (violet). a b c d 245
JKSMRM 16(3) : 243-252, 2012 (V 1 - V 2 ) 또한새로운정규분포를형성한다고추론할수있다. 이러한정규분포의평균과분산은 Eq. [2] 및 Eq. [3] 으로정의된다. 동일인의중복측정에대한체적을고려하면평균은 0 분산은 2 ( = V 1 = V 2 ) 로가정할수있고양측검정 α<0.05 에서의표준정규분포화를실시하여 Eq. [4] 와같이정리할수있다. V 1 - V 2 PVD = 100 [1] V 1 - V 2 = V1 - V 2 [2] V 1 - V2 = V 1 2 + V2 2 [3] X V1 -V 2 > V1 -V 2 +Z 2 2 X V1 -V 2 - V1 - V 2 > Z 2 2 X V1 -V 2 > 2.772 PVD > 2.772(CV) [4] 결 과 V 1 + V 2 2 Table 2 는 4 개의 MRI 장치에서획득한동일대상자의자동분할된각영역별체적에대한통계적일치성 (inter MR scanner reliability, ICC) 및변이성 (inter MR scanner variability, CV) 의평균을나타낸다. ICC 를이용한 MRI 장치간모든영역에대한체적일치성은평균 0.837 (± 0.082) 이며, lateral ventricle 에서 0.995 (0.988-999) 로가장높은일치성을나타내었고, 왼쪽 amygdala 에서 0.691 (0.410-0.897) 로가장낮은일치성을보였다. 상대적으로 GM, lateral ventricle 등의체적이큰영역에서높은일치성이확인되었다. CV 를이용한 MRI 장치간의모든영역에대한체적변이성은평균 4.176% (±1.724) 이며, GM 에서 2.175% (±1.466) 로가장낮은변이성을나타내었고, 오른쪽 amygdala 에서 5.972% (±2.119) 로가장높은변이성을보였다. 통계적으로왼쪽 amygdala 의체적이 MR(a) vs. MR(c), MR(b) vs. MR(c), MR(c) vs. MR(d) 에서유의한차이가있었고, 오른쪽 caudate 에서는 MR(b) 와 MR(c) 에서차이가있었다. 오른쪽 hippocampus 에서는 MR(c) 가다른 MR 장치와비교하여모두유의한차이가있었고, 왼쪽 putamen 과오른쪽 thalamus 에서는 MR(d) 가다른 MR 장치들과차이가있었다. 오른쪽 putamen 은 MR(b) vs. MR(d), 왼쪽 thalamus 는 MR(b) vs. MR(c) 에서각각차이를나타내었다. Figure 2 는 Eq. [1] 과같이정의된 PVD 를이용하여동일대상자에대해각각의 MRI 장치사이에서 (e.g. MR(a) vs. MR(b)) 발생된각영역별체적의차이를나타낸다. MR(a) 와 (b) 는각각의 Philips 3T 장치에서획득한영상, MR(c) Table 2. Inter MR Scanner Reliability and Variability for Segmentation Volumes Region ICC 1) CV (%) 2) RM-ANOVA 3) GM 4) 0.952 (0.881-0.986) 2.175±1.466 0.069 Amyg L 0.691 (0.410-0.897) 5.655±2.390 0.009**,#,+ Amyg R 0.782 (0.547-0.931) 5.972±2.119 0.204 Caud L 0.768 (0.524-0.926) 4.509±2.955 0.787 Caud R 0.816 (0.605-0.943) 4.229±2.955 0.035 # Hipp L 0.845 (0.657-0.953) 3.112±1.282 0.283 Hipp R 0.838 (0.645-0.951) 3.692±1.325 0.018**,#,+ Put L 0.773 (0.532-0.928) 4.661±1.462 0.005***,##,+ Put R 0.862 (0.689-0.958) 3.786±1.232 0.003 ## Thal L 0.849 (0.664-0.954) 5.004±2.142 0.028 # Thal R 0.872 (0.708-0.962) 5.016±1.587 0.001***,##,+ LV 0.995 (0.988-0.999) 3.907±1.694 0.230 All regions 5) 0.837±0.082 4.310±1.879 Note: Maximum and minimum values of each column are indicated with a bold font 1) ICC: intraclass correlation coefficient, data are statistics with 95% confidence interval 2) CV: coefficient of variation, data are mean±standard deviation derived from all subjects 3) RM-ANOVA: repeated measures ANOVA, data are statistics of significance level (p-value) *, **, *** : significant difference (p < 0.05) between MR(a) and others (MR(b), MR(c), MR(d)) #, ## : significant difference (p < 0.05) between MR(b) and others (MR(c), MD(d)) + : significant difference (p < 0.05) between MR(c) and MD(d) 4) GM: gray matter, Amyg: amygdala, Caud: caudate nucleus, Hipp: hippocampus, Put: putamen, Thal: thalamus, LV: lateral ventricle, L: left, R: right 5) All regions in first column represent mean±standard deviation on ICC and that of second column show the average in mean values and standard deviation values on Mean CV 246
다기관의자기공명영상장치에서측정된뇌조직체적의재현성분석 정원범외 는 Siemens 3T 장치에서획득한영상, MR(d) 는 GE 1.5T 에서획득한영상을나타낸다. 전체적으로각영역에서같은장비 MR(a) vs. MR(b) 사이에서발생하는체적의차가평균 3.611% (±2.774) 로가장낮았고, 주자장강도는같으나제조사및영상획득인자가다른 MRI 장치사이에서 (i.e. MR (a) vs. MR (c), MR (b) vs. MR (c)) 전체영역에대해각각 5.623%(±4.121), 5.093%(±3.631) 의체적차이가발생하였다. 주자장강도및제조사그리고영상획득인자가다른 MRI 장치사이에서 (i.e. MR (a) vs. MR (d), MR (b) vs. MR (d), MR (c) vs. MR (d)) 전체영역에대해각각 5.035%(±3.766), 5.923%(±3.712), 6.398%(±4.434) 의체적차이가확인되었으며, MR (c) vs. MR (d) 사이에서발생하는체적의차가가장가장컸다. 영역에따른체적의차이는 GM 에서평균 2.697% (± 2.267) 로가장작았으며, 오른쪽 amygdala 에서평균 7.252 (±4.993) 로가장컸다. Table 3 은각 MRI 장치에서획득된영상을이용하여장치간특성의차이로인해발생되는체적의차이를감안하고질환의영향으로인한뇌구조의변화를가늠할경우각영역에대해서요구되는최소한의체적오프셋을나타낸다. Eq. [4] 와같이정의된수식을이용하여동일대상자에대해각각의 MRI 장치사이에서발생된체적의차이를기반으로추정하였으며, 체적의차가이이상의범위일경우통계적으로유의한체적차이가있는것으로가정할수있다 Table 3. Thresholds of Significant Percent Volume Difference (α= 0.05, two-tailed) Between Two MR Derived Volumetric Results Region MR(a) MR(b) MR(a) MR(c) MR(a) MR(d) MR(b) MR(c) MR(b) MR(d) MR(c) MR(d) GM 1) 3.047 6.611 3.113 6.018 4.882 8.050 Amyg L 13.190 19.165 9.483 11.645 11.252 17.893 Amyg R 10.971 18.590 9.568 13.079 14.319 18.755 Caud L 5.289 12.703 13.300 11.440 12.961 10.197 Caud R 4.866 8.914 13.081 9.647 14.955 10.662 Hipp L 5.930 8.758 4.600 9.547 6.158 9.466 Hipp R 6.431 10.861 4.594 12.175 6.026 11.335 Put L 9.178 9.779 12.404 9.632 16.039 11.261 Put R 6.696 7.257 12.025 7.462 11.305 10.099 Thal L 6.950 11.751 10.702 13.584 13.056 15.828 Thal R 5.170 7.521 18.148 6.495 18.451 14.141 LV 7.229 10.117 7.413 9.070 9.910 12.811 All regions 2) 7.079 11.022 9.870 9.983 11.609 12.541 Note: Unless otherwise indicated, data are mean values (%) from all subjects Maximum and minimum values of each column are indicated with a bold font 1) GM: gray matter, Amyg: amygdala, Caud: caudate nucleus, Hipp: hippocampus, Put: putamen, Thal: thalamus, LV: lateral ventricle, L: left, R: right 2) All regions represent the mean values of each column Fig. 2. Percent volume differences between two MR derived volumetric results for brain structures segmented with automatic processing 247
JKSMRM 16(3) : 243-252, 2012 (α= 0.05, two-tailed). MR(a) vs. MR(b) 에서는 GM 이최소 3.047% 이상의체적차, 왼쪽 amygdala 에서 13.190% 로가장큰체적차이가나타났다. MR(a) vs. MR(c) 에서는 GM 6.611%, 왼쪽 amygdala 19.165% 로최소 최대범위를각각나타낸다. MR(a) vs. MR(d), MR(b) vs. MR(c), MR(b) vs. MR(d) 에서는 GM 가각각 3.313%, 6.018%, 4.882% 로가장작은체적의차, thalamus 에서각각 18.148% (right), 13.584% (left), 18.451% (right) 의가장높은체적차오프셋이요구되었으며, MR(c) vs. MR(d) 영상을비교할경우에는 GM 에서최소 8.050% 이상, 오른쪽 amygdala 에서최소 18.755% 로가장높은체적차오프셋이필요한것을확인하였다. 결 론 본연구에서는 MRI 를이용한뇌질환진단및분석을목적으로하는다기관연구환경구축을위한기초연구로서자동화분할 S/W 를이용하여다른 MRI 장치간에서발생된체적차이를측정하고, 통계적유의수준에서의장치간에측정된뇌조직의체적차이를확인하기위한최소범위를확인하였다. MRI 를이용한뇌조직구조의병리학적변화는질환의상태및병인과관련된영상마커에서의유효한체적감소로서확인된다. 대표적인퇴행성뇌질환으로서알츠하이머성치매는인지및기억기능의감퇴와관련하여 hippocampus, amygdala, enthorhinal cortex 및 posterior cingulate cortex 등의해마신경로 (hippocampal pathway) 에서두드러진위축이발생하며 (36), 점차적으로언어, 행동및시공간적기능의장애와함께전반적인대뇌피질에서의변화가나타난다 (37). 또한 putamen 과 thalamus 에서도인지기능과유의한상관성을보이며체적이감소하며 (38), 체적의감소와상응하여 lateral ventricle 에서는노화에따라년간 1.5~3.0% 의체적증가에비해알츠하이머성치매의경우, 약 5~16% 증가하는것으로알려져있다 (39, 40). 또다른퇴행성뇌질환으로서파킨슨병은불안정한움직임및운동장애증상 (movement disorder) 과관련하여전체적인 motor 기능을조절하는 putamen (9), 자발적인움직임및기억과학습능력을조절하는 caudate nucleus (41), 운동기능을억제또는촉진하는 thalamus (42) 등의기저핵 (basal ganglia) 에서의두드러진변화가나타난다. MR 영상을이용한헌팅턴병에대한분석결과, 파킨슨병과유사하게전체적인뇌체적의감소와함께 caudate nucleus 와 putamen 에서유의한위축이있는것으로확인된다 (14). 정신분열증환자의경우, 정상인과비교하여, amygdala 와 hippocampus 를포함한 amygdala-hippocampal complex 영역과 putamen 및 thalamus 에서체적감소가 있는것으로보고된다 (43-45). 또다른정신질환인우울증과조울증에대해, 다른뇌영역에비해 hippocampus 와 amygdala 등의영역에서체적감소가있는것으로확인된다 (46-48). 통계적검증력확보를위한다기관연구에서는질환으로인한뇌조직의변화를정확하게감지하기위해 MRI 장치간에발생하는오차및최소변화범위를확인하여야한다. 본연구에서는앞서언급한연구결과를토대로 Figure 1 과같이관심영역을설정하여 4 개의 MRI 장치에서획득된동일인에대한체적의일치도및차이를확인하였다. 통계적으로결과의일치도및신뢰성을나타내는 ICC 는 amygdala 에서최소 0.691, lateral ventricle 에서최대 0.995 이며, 모든영역에대한평균은 0.837 의결과를확인하였다 (Table 2). 일반적으로 ICC 결과가 0.75 이상일경우신뢰할수있는결과를의미한다 (49-51). 체적결과의오차범위로 GM 에서최소 2.175%, amygdala 에서최대 5.972%, 모든영역에대해평균 4.310% 의 CV 결과를확인하였다. 전체적인일치도및체적차이에대해 amygdala 의결과가가장신뢰성이낮았으며, 이에대해 MR 영상의대조도 (contrast) 가비슷하고 (30), 위치상으로인접한 hippocampus 를함께포함하여 (amygdala-hippocampal complex) 체적을비교할경우, 좌 0.884 (0.733-0.966) 우 0.890 (0.745-0.967) 의 ICC 및좌 3.114% (± 0.643) 우 3.653% (±0.950) 의 CV 가확인되었다. 이는기존의 amygdala 및 hippocampus 개별영역에대한분석보다향상된수치이며, 질환에따른체적의감소가 amygdala 와 hippocampus 에서유사하게발생되는점과구조적영상에서두영역간위치및경계가모호한점을고려하여본연구에서는두영역을함께포함하여체적을분석하는방법이효과적인것으로제안된다. 주자장의세기및영상획득인자등시스템적으로다른 MRI 장치간의체적분석에대한이전의유사연구에서는대상자 1 명에대해각영역별 CV 계수를이용하여체적을비교한결과최소 0.66%, 최대 14.71% 및평균약 4.74% (median value) 의체적변화를나타내었다 (52). 또다른연구에서는전체뇌 (total brain) 체적에대해 0.80 의 ICC 가확인되었으나, 신호강도히스토그램을조절하였을경우 0.96 으로향상되었으며 (53), 이를미루어보아시스템적차이에따른체적의변화에대해소프트웨어적조절이필요함을알수있다. 4 개의전체 MRI 장치에서의체적변화 (CV) 와더불어각시스템간의체적차이를 (PVD) 고려하였을때, 시스템특성및영상획득인자가같을경우 MR 장치간의 (MR(a) vs. MR(b)) 체적변화가전체영역에대해평균 3.611% (± 2.774) 로가장작았다. 약한달이내에두시스템에서영상이획득됨에따라정상인에대해시간에따른뇌조직의체적변화가매우작을것이라고감안할때, 장치의제조사, 주자장세기, 영상획득인자가같더라도발생되는시스템간의체적변화는 (inter-site variation) 주자장의안정성 248
다기관의자기공명영상장치에서측정된뇌조직체적의재현성분석 정원범외 (magnetic inhomogeneity) 및비선형적인 gradient 세기등의시스템적오차 (systemic error) 에따른기하학적왜곡 (geometric distortions) 과대상자의움직임 (motion artifact) 등의영향인것으로여겨진다. 영상촬영간미세한움직임은부분체적효과 (partial volume effect) 를발생시켜다른조직간의 (different tissue types) 구분이나인접한하부구조간의경계를모호하게하는등영상대조도에영향을끼침에따라자동화영역분할방법을이용할경우, 부정확한결과를초래할수있다. 주자장강도는같고, 제조사및영상획득인자가다른 MRI 장치사이에서전체영역에대해평균 5.358% 의체적차이가있었고, 주자장강도및제조사, 영상획득인자가다른 MRI 장치사이에서전체영역에대해평균 5.785% 의체적차이가발생되었다. 이는앞서언급된시스템적오차및움직임에대한영향과더불어장치의차이에따른영향으로발생된것이며, 위의결과를통해주자장의강도가같을경우발생되는체적의차이가더작음을확인할수있다. 각영역에대한시스템간최소및최대체적차이는 (Figure. 2) 각각 GM 에서평균 2.697% (±2.267), amygdala 에서평균 7.252% (±4.993) 였다. 이를기반으로 MRI 장치간유효한체적차이를 (α=0.05, two-tailed) 나타내기위한최소한의체적차오프셋을확인하였으며 (Table 3), 이이상의범위일경우, 각 MRI 장치에서획득된그룹간의 (controls vs. patients) 통계적으로유효한체적의변화가있는것으로간주할수있다. 앞서 CV 를이용한체적변화의분석방법과같이 amygdala 와 hippocampus 영역을함께포함하여체적차이를확인하였을경우, 좌평균 3.818% (±2.082) 우평균 4.416% (±2.685) 로차이가감소하였다. 시간간격을두고동일한 MRI 장치및영상획득인자를통해획득된영상을본연구와같은자동화분할 S/W 을이용하여체적을분석한이전의연구에서는 (scan-rescan variation) 각영역에대해평균 3.2% (±2.783) 의체적차이 (PVD) 가있었고 (54), 또다른연구에서는영상획득방향에따라회백질및백질영역에서약 1%, 해마영역에서약 3% 의체적차이 (CV) 가확인되었다 (55). 이를통해다기관연구구축을위한 MRI 장치간의체적차이와더불어장기적연구를 (longitudinal study) 위해서는 MRI 장치내에서발생되는체적의차이또한고려되어야함을알수있다. MR 영상분석을이용한다기관연구기반의통계적결론은각기관에서정립한영상처리프로토콜에따라분석된결과를취합하여정량적해석을수행하거나 (56,57), 각기관에서의영상을하나의기관에서취합하여정립된영상처리프로토콜에맞춰분석할수있다. 그러나각기관에서분석한결과를토대로통계적결론을도출할경우분석방법의차이로인해발생되는해석적변동성이 (site variation) 있으며, 본연구와같이통상의 PC 를사용하여자동화분할 S/W 를 (FreeSurfer) 사용했을때한영상당약 7 시간의분석시간이소요됨을고려할때, 하나의기관에 249 서분석할경우데이터수집및분석에대한시간적효율이떨어질수있는단점을가진다. 본연구에서는다기관연구구축을위해시스템적으로다른 MRI 장치에서획득된동일인의영상을자동분할하여각영역별체적의차이를분석하고, 시스템간의발생되는통계적유효체적범위를측정하였으며, 이를기반으로질환에따른뇌체적의변화를추정할수있는유의성을확인하였다. 추가적으로본연구에서의제한점을토대로대량의데이터수집을위한데이터베이스마련및 GPU 기반의병렬처리를 (GPU based parallel processing) 적용한고속영상처리방법등이요구되며이를일률적으로관리할수있는시스템 (frame work) 개발또한필요할것임을확인하였다. 참고문헌 1. Basser PJ, Pierpaoli C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. Journal of Magnetic Resonance-Series B 1996;111:209-219 2. Voss HU, Schiff ND. MRI of neuronal network structure, function, and plasticity. Progress in brain research 2009;175:483-496 3. Whitwell JL. Voxel-based morphometry: an automated technique for assessing structural changes in the brain. The Journal of Neuroscience 2009;29:9661-9664 4. Um M, Park B, Park HJ. Anatomical Brain Connectivity Map of Korean Children. J Korean Soc Magn Reson Med 2011;15:110-122 5. Scahill RI, Frost C, Jenkins R, Whitwell JL, Rossor MN, Fox NC. A longitudinal study of brain volume changes in normal aging using serial registered magnetic resonance imaging. Archives of Neurology 2003;60:989-994 6. Ge Y, Grossman RI, Babb JS, Rabin ML, Mannon LJ, Kolson DL. Age-related total gray matter and white matter changes in normal adult brain. Part I: volumetric MR imaging analysis. American Journal of Neuroradiology 2002;23:1327-1333 7. Choi S, Kim WY, Lee KN, et al. The age-related microstructural changes of the cortical gray and white matter ratios on T2-, FLAIR and T1-weighted MR images. J Korean Soc Magn Reson Med 2011;15:32-40 8. Guo X, Wang Z, Li K, et al. Voxel-based assessment of gray and white matter volumes in Alzheimer s disease. Neuroscience letters 2010;468:146-150 9. Geng D, Li YX, Zee CS. Magnetic resonance imaging-based volumetric analysis of basal ganglia nuclei and substantia nigra in patients with Parkinson s disease. Neurosurgery 2006;58: 256-262 10. Mascalchi M, Lolli F, Della Nave R, et al. Huntington disease: volumetric, diffusion-weighted, and magnetization transfer MR imaging of brain1. Radiology 2004;232:867-873 11.Mathalon DH, Sullivan EV, Lim KO, Pfefferbaum A. Progressive brain volume changes and the clinical course of schizophrenia in men: a longitudinal magnetic resonance imaging study. Archives of General Psychiatry 2001;58:148-157 12. Koolschijn P, Van Haren NEM, Lensvelt-Mulders GJLM, Hulshoff Pol HE, Kahn RS. Brain volume abnormalities in major
JKSMRM 16(3) : 243-252, 2012 depressive disorder: a meta-analysis of magnetic resonance imaging studies. Human Brain Mapping 2009;30:3719-3735 13. Blumberg HP, Kaufman J, Martin A, et al. Amygdala and hippocampal volumes in adolescents and adults with bipolar disorder. Archives of General Psychiatry 2003;60:1201-1208 14. Mascalchi M, Lolli F, Della Nave R, et al. Huntington disease: volumetric, diffusion-weighted, and magnetization transfer MR imaging of brain. Radiology 2004;232:867 15.Pievani M, De Haan W, Wu T, Seeley WW, Frisoni GB. Functional network disruption in the degenerative dementias. The Lancet Neurology 2011;10:829-843 16. van de Pol LA, Korf ESC, van der Flier WM, et al. Magnetic resonance imaging predictors of cognition in mild cognitive impairment. Archives of Neurology 2007;64:1023 17. Fazekas F, Kapeller P, Schmidt R, Offenbacher H, Payer F, Fazekas G. The relation of cerebral magnetic resonance signal hyperintensities to Alzheimer s disease. Journal of the Neurological Sciences 1996;142:121-125 18. Schmidt R, Schmidt H, Kapeller P, et al. The natural course of MRI white matter hyperintensities. Journal of the Neurological Sciences 2002;203:253-257 19. Reiss AL, Eckert MA, Rose FE, et al. An experiment of nature: brain anatomy parallels cognition and behavior in Williams syndrome. The Journal of Neuroscience 2004;24:5009-5015 20. Eckert MA, Tenforde A, Galaburda AM, et al. To modulate or not to modulate: differing results in uniquely shaped Williams syndrome brains. Neuroimage 2006;32:1001-1007 21. Zhu T, Hu R, Qiu X, et al. Quantification of accuracy and precision of multi-center DTI measurements: a diffusion phantom and human brain study. Neuroimage 2011;56:1398-1411 22.Suckling J, Ohlssen D, Andrew C, et al. Components of variance in a multicentre functional MRI study and implications for calculation of statistical power. Human Brain Mapping 2008;29:1111-1122 23. Suckling J, Barnes A, Job D, et al. Power calculations for multicenter imaging studies controlled by the false discovery rate. Human Brain Mapping 2010;31:1183-1195 24. Jack Jr CR, Bernstein MA, Fox NC, et al. The Alzheimer s disease neuroimaging initiative (ADNI): MRI methods. Journal of Magnetic Resonance Imaging 2008;27:685-691 25. Moorhead TW, Gountouna VE, Job D, et al. Prospective multicentre voxel based morphometry study employing scanner specific segmentations: procedure development using CaliBrain structural MRI data. BMC Medical Imaging 2009;9:8 26. Balamoody S, Williams T, Waterton J, et al. Comparison of 3T MR scanners in regional cartilage-thickness analysis in osteoarthritis: a cross-sectional multicenter, multivendor study. Arthritis Research and Therapy 2010;12:202 27. Takao H, Hayashi N, Ohtomo K. Effect of scanner in longitudinal studies of brain volume changes. Journal of Magnetic Resonance Imaging 2011;34:438-444 28. Friston KJ, Holmes AP, Worsley KJ, Poline JP, Frith CD, Frackowiak RSJ. Statistical parametric maps in functional imaging: a general linear approach. Human Brain Mapping 1994;2:189-210 29. May A, Gaser C. Magnetic resonance-based morphometry: a window into structural plasticity of the brain. Current Opinion in Neurology 2006;19:407 30. Fischl B, Salat DH, Busa E, et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 2002;33:341-355 31. Fischl B, Salat DH, van der Kouwe AJW, et al. Sequenceindependent segmentation of magnetic resonance images. NeuroImage 2004;23:69-84 32. Cocosco CA, Kollokian V, Pike GB, Evans AC. BrainWeb: online interface to a 3D MRI simulated brain database. Neuroimage 1997;5:425 33. Klauschen F, Goldman A, Barra V, Meyer-Lindenberg A, Lundervold A. Evaluation of automated brain MR image segmentation and volumetry methods. Human Brain Mapping 2009;30:1310-1327 34. Morey RA, Petty CM, Xu Y, Pannu Hayes J, Wagner II HR, Lewis DV. A comparison of automated segmentation and manual tracing for quantifying hippocampal and amygdala volumes. Neuroimage 2009;45:855-866 35. Pardoe HR, Pell GS, Abbott DF, Jackson GD. Hippocampal volume assessment in temporal lobe epilepsy: how good is automated segmentation? Epilepsia 2009;50:2586-2592 36.Scahill RI, Schott JM, Stevens JM, Rossor MN, Fox NC. Mapping the evolution of regional atrophy in Alzheimer s disease: unbiased analysis of fluid-registered serial MRI. Proceedings of the National Academy of Sciences 2002;99:4703 37. McDonald C, McEvoy L, Gharapetian L, et al. Regional rates of neocortical atrophy from normal aging to early Alzheimer disease. Neurology 2009;73(6):457-465 38. De Jong L, Van Der Hiele K, Veer I, et al. Strongly reduced volumes of putamen and thalamus in Alzheimer s disease: an MRI study. Brain 2008;131:3277-3285 39. Shear PK, Sullivan EV, Mathalon DH, et al. Longitudinal volumetric computed tomographic analysis of regional brain changes in normal aging and Alzheimer s disease. Archives of Neurology 1995;52:392 40. Sullivan E, Pfefferbaum A, Adalsteinsson E, Swan G, Carmelli D. Differential rates of regional brain change in callosal and ventricular size: a 4-year longitudinal MRI study of elderly men. Cerebral Cortex 2002;12:438-445 41. Lee S, Kim S, Tae W, et al. Regional volume analysis of the Parkinson disease brain in early disease stage: gray matter, white matter, striatum, and thalamus. American Journal of Neuroradiology 2011;32:682-687 42. Asami T, Bouix S, Whitford TJ, Shenton ME, Salisbury DF, McCarley RW. Longitudinal loss of gray matter volume in patients with first-episode schizophrenia: DARTEL automated analysis and ROI validation. Neuroimage 2011;59:986-996 43. Seidman LJ, Faraone SV, Goldstein JM, et al. Thalamic and amygdala-hippocampal volume reductions in first-degree relatives of patients with schizophrenia: an MRI-based morphometric analysis. Biological Psychiatry 1999;46:941-954 44. Kreczmanski P, Heinsen H, Mantua V, et al. Volume, neuron density and total neuron number in five subcortical regions in schizophrenia. Brain 2007;130:678-692 45. Gur RE, Turetsky BI, Cowell PE, et al. Temporolimbic volume reductions in schizophrenia. Archives of General Psychiatry 2000;57:769-775 46. Bremner JD, Narayan M, Anderson ER, Staib LH, Miller HL, Charney DS. Hippocampal volume reduction in major depression. American Journal of Psychiatry 2000;157:115-118 47. Sapolsky RM. Depression, antidepressants, and the shrinking hippocampus. Proceedings of the National Academy of Sciences 2001;98:12320 48.Chang K, Karchemskiy A, Barnea-Goraly N, Garrett A, 250
다기관의자기공명영상장치에서측정된뇌조직체적의재현성분석 정원범외 Simeonova DI, Reiss A. Reduced amygdalar gray matter volume in familial pediatric bipolar disorder. Journal of the American Academy of Child & Adolescent Psychiatry 2005;44:565-573 49. Brooks D, Thomas J. Interrater reliability of auscultation of breath sounds among physical therapists. Physical Therapy 1995;75:1082-1088 50. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychological Bulletin 1979;86:420-428 51. Fleiss JL. The measurement of interrater agreement. Statistical methods for rates and proportions 1981;2:212-236 52. Huppertz HJ, Kröll-Seger J, Klöppel S, Ganz RE, Kassubek J. Intra-and interscanner variability of automated voxel-based volumetry based on a 3D probabilistic atlas of human cerebral structures. Neuroimage 2010;49:2216-2224 53.Schnack HG, van Haren NEM, Hulshoff Pol HE, et al. Reliability of brain volumes from multicenter MRI acquisition: a calibration study. Human Brain Mapping 2004;22:312-320 54. Morey RA, Selgrade ES, Wagner II HR, Huettel SA, Wang L, McCarthy G. Scan-rescan reliability of subcortical brain volumes derived from automated segmentation. Human Brain Mapping 2010;31:1751-1762 55. Jung WB, Son DB, Kim YJ, Kim YH, Eun CK, Mun CW. A comparison study on human brain volume of white matter, gray matter and hippocampus depending on magnetic resonance imaging conditions and applied brain template. J Korean Soc Magn Reson Med 2011;15:242-250 56. Campbell S, Marriott M, Nahmias C, MacQueen GM. Lower hippocampal volume in patients suffering from depression: a meta-analysis. American Journal of Psychiatry 2004;161:598-607 57. Videbech P, Ravnkilde B. Hippocampal volume and depression: a meta-analysis of MRI studies. American Journal of Psychiatry 2004;161:1957-1966 251
JKSMRM 16(3) : 243-252, 2012 JKSMRM 16(3) : 243-252, 2012 Reproducibility Analysis of Brain Volumetry Measured from Inter MR Scanner of Multi-Institute Won Beom Jung 1, Min Jae Kang 1, Doo Beom Son 2, Young Joo Kim 2, Young Min Lee 3, Young Hoon Kim 4, Choong Ki Eun 5, Chi Woong Mun 1 1 Department of Biomedical Engineering, and U-Health care Research Center, Inje University 2 Department of Diagnostic Radiology, Haeundae Paik Hospital 3 Department of Psychiatry, Medical School, Pusan National University, Pusan National University Hospital 4 Department of Psychiatry, Medical School, Inje University, Haeundae Paik Hospital 5 Department of Diagnostic Radiology, Medical School, Inje University, Haeundae Paik Hospital Purpose : The aim of this study was to evaluate the variations of brain volumetry between the different MR scanners or the different institutes. Materials and Methods: Ten normal subjects were scanned at four different MR scanners, two of them were the same models, to measure inter-mr scanner variations using intraclass correlation coefficient (ICC), coefficient of variation (CV) and percent volume difference (PVD) and to calculate minimal thresholds to detect the significant volumetric changes in gray matter and subcortical regions. Results: Averaged statistical reliability (ICC = 0.837) and volumetric variation (CV = 4.310%) in all segmented regions were observed on overall MR scanners. Comparing the segmented volumes with PVD between two MR scanners, volumetric differences on same models were the lowest (PVD = 3.611%) and volume thresholds were calculated with 7.168%. PVD results and thresholds values on systemically different MR scanners were evaluated with 5.785% and 11.340% respectively. Conclusion: Authors conclude that the reliability of brain volumetry is not so high. Calibration studies of MRI system and image processing are essential to reduce the volumetric variability. Additionally, frameworks comprised of database and algorithms with high-speed image processing are also required for the efficient image data management. Index words : Multi-center study Inter MR scanner variation Brain volumetry Magnetic resonance image Automatic segmentation Address reprint requests to: Chi Woong Mun, Ph.D., Department of Biomedical Engineering and UHRC, Inje University 607, Obang-dong, Gimhae-si, Gyeongnam 621-749, Korea. Tel. 82-55-320-3297 Fax. 82-55-327-3292 E-mail: mcw@inje.ac.kr 252