View : 439 Download: 0

Full metadata record

DC Field Value Language
dc.description.abstractThe most strong point of positron emission tomography (PET) is the quantitative estimate of physiological parameters such as blood flow and glucose metabolism. To measure these parameters quantitatively, it is very important to get exact time-activity curves (TACs) of blood pool and tissue. In cardiac study, region of interest (ROI) method has been usually used to obtain TACs from PET image or arterial blood was taken for blood pool TAC. In brain study, arterial blood sampling should be performed for blood pool TAC because brain PET images do not include left ventricle that is used as blood pool. Both ROI method and arterial sampling, however, have some problems that can cause error for the quantitative estimates. Especially in H_(2)^(15)O PET study, ROI method is not appropriate to get TACs due to low image quality. For the quantitative estimation of regional myocardial blood flow (rMBF), C^(15)O blood pool scan should be performed to draw ROI. In this dissertation, to overcome such problems factor analysis and independent component analysis (ICA) were applied and pure TACs of left ventricle and tissue were extracted from myocardial H_(2)^(15)O PET and pure TAC of carotid artery in brain H_(2)^(15)O PET. The purpose of both methods is separating blind sources from observed data. ICA, especially, not only de-correlates the signals but also reduces higher-order statistical dependencies. Therefore ICA is expected to separate independent components even from low quality images. Factor analysis was separated left ventricle, right ventricle, and myocardium from myocardial H_(2)^(15)O PET images and rMBF estimated with TACs by factor analysis had good correlation with rMBF by microsphere. ICA was successfully extracted carotid arterial component from brain H_(2)^(15)O PET images and TAC of carotid artery was nearly same to that of arterial sampling. Both factor analysis and ICA will be helpful for the quantitative analysis in dynamic H_(2)^(15)O PET images. ;양전자단층촬영(PET)의 가장 큰 장점은 혈류나 포도당대사등의 생리학적 지표를 비침습적이고 정량적으로 계산해낼 수 있다는 것이다. 이러한 지표들을 정확히 계산하기 위해서는 우선 혈액풀(blood pool)과 조직의 시간-방사능곡선(TAC)을 정확하게 찾는 것이 매우 중요하다. TAC는 시간에 따른 방사능 농도의 양으로 기존의 연구에서는 관심영역(ROI)을 그려 TAC를 얻거나 동맥혈을 직접 취하여 혈액풀 TAC로 사용해왔다. ROI를 직접 그리기 어려운 경우 심근 H_(2)^(15)O PET 연구에서는 추가적인 C^(15)O PET 촬영이 필요했으며, 뇌 PET 연구에서는 영상에 좌심실이 포함되지 않아 동맥혈을 채취하여 혈액풀 TAC로 사용하였다. 그러나 ROI 방법은 사용자 의존적이며 특히 영상의 질이 떨어질 경우에는 ROI를 정확히 그리기 어려워 결과에 심각한 영향을 미칠 수 있다. 또한 동맥혈을 채취하는 방법은 환자에게 고통을 줄 수 있으므로 바람직한 방법이 되지 못한다. 이 연구에서는 이러한 문제점들을 극복하고자 심근 H_(2)^(15)O PET 영상에 인자분석(factor analysis)을 적용하여 순수한 좌심실 TAC와 조직 TAC를 분리해내고 이로부터 국소심근혈류(rMBF)를 계산하고자 하였다. 또한 뇌 H_(2)^(15)O PET 영상에 독립성분분석법(independent component analysis; ICA)을 적용하여 영상으로부터 직접 경동맥 TAC를 얻고자 하였다. 이 두 방법의 목적은 관측한 데이터들로부터 숨겨진 독립성분들을 분리해 내는 것이다. 특히 ICA는 신호들의 상관관계를 줄일 뿐 아니라 고차통계적독립성(higher-order statistical dependencies)을 줄여가며 각각의 독립성분들을 분리해 내는 것으로 최근 의학영상 분석에 큰 주목을 받는 방법이다. 본 연구에서 고안한 두 단계 인자분석을 통하여 심근 H_(2)^(15)O PET 영상으로부터 추가적인 PET 촬영 없이 rMBF를 비침습적이고 정량적으로 계산하였다. 또한 뇌 H_(2)^(15)O PET 영상에 ICA를 적용하여 동맥혈을 채취하지 않고 성공적으로 경동맥 TAC를 찾아 내었다. 따라서 인자분석과 ICA는 H_(2)^(15)O PET 영상의 정량적 분석에 큰 도움이 되리라 기대된다.-
dc.description.tableofcontentsTable of Contents = i List of Figures = vi List of Tables = ix List of Abbreviations = x Terminology of Medicine = xi Acknowledgement = xii Abstract = xiv I. Introduction = 1 A. Positron Emission Tomography = 3 1. Principle of PET = 3 2. Radioisotope for PET scan = 7 B. Tracer Kinetic Models for Blood Flow Measurement = 9 1. Microsphere and tracer trapping models = 10 2. Single compartment model = 12 3. Input function and tissue time-activity curve = 15 4. Kinetic model for H_(2)^(15)O: Correction of partial volume and spillover effect = 17 II. Quantification of Regional Myocardial Blood Flow Using Dynamic H_(2)^(15)O PET and Factor Analysis = 19 A. Introduction = 19 B. Materials and Methods = 22 1. Factor analysis: Theory = 22 2. Image acquisition and reconstruction = 25 3. Microsphere study = 26 4. Image preprocessing = 27 5. Factor analysis for H_(2)^(15)O dynamic PET image = 28 6. Comparison of input functions derived by factor analysis and arterial sampling = 32 7. Calculation of rMBF by factor analysis and ROI method = 32 8. Data analysis = 33 C. Results = 34 1. Factor analysis = 34 2. rMBF calculation and validation = 37 D. Discussion = 42 III. Derivation of Input Function in Carotid Artery from Brain H_(2)^(15)O PET images = 48 A. Introduction = 48 B. Materials and Methods = 51 1. Independent component analysis (ICA) = 51 2. Image acquisition and reconstruction = 55 3. Image preprocessing = 57 4. Input function derivation using ICA and factor analysis = 57 5. Validation of input functions by ICA and factor analysis = 60 6. Human study = 61 C. Results = 61 1. H_(2)^(15)O PET study = 61 2. F-18-FDG PET study: Human study = 66 D. Discussion = 68 IV. Conclusions = 72 V. References = 75 VI. Appendix = 87 A. Factor Analysis for Dynamic PET Image = 87 1. The algorithm for factor analysis = 88 B. Information Maximization; Informax = 93 국문초록 = 101 List of publication = 103-
dc.format.extent1292437 bytes-
dc.publisher이화여자대학교 대학원-
dc.titleAnalysis of Dynamic H₂15O PET Data by Factor Analysis and Independent Component Analysis-
dc.typeDoctoral Thesis-
dc.format.pagevx, 107 leaves-
dc.identifier.major대학원 물리학과- 8-
Appears in Collections:
일반대학원 > 물리학과 > Theses_Ph.D
Files in This Item:
There are no files associated with this item.
RIS (EndNote)
XLS (Excel)


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.