Modeling and Correction of Event Localization Uncertainties in High-Resolution PET Detectors : Focusing on Depth of Interaction and Inter Crystal Scattering = 상호작용 깊이 및 결정 간 산란을 중심으로 한 고해상도 PET 검출기에서의 이벤트 위치 불확실성 모델링 및 보정

국문 초록 (Abstract)

본 논문은 고해상도 PET 검출기에서의 이벤트 위치 불확실성(event-localization uncertainty)을 분석하고 보정하기 위한 통합적 연구를 제시한다. 특히, multiplexed side-readout 구조를 기반으로 하여 광전 신호의 손실 없이 위치 정보를 복원할 수 있는 방법을 체계적으로 구축하였다. 연구의 주요 내용은 (1) inter-crystal scatter (ICS) 이벤트의 판별 및 위치 복원, (2) 연속형 및 분절형 결정에서의 DOI (Depth-of-Interaction) 재구성, (3) 광학적 confinement 및 분절 피치 변화가 DOI 성능에 미치는 영향 분석으로 구성된다.
본 연구에서는 CPLD 기반 multiplexed front-end 회로, 인공지능 기반 DOI 복원 알고리즘(ANN), 그리고 회귀 기반 generalized proportional 모델을 제안하였다. ANN 기반 DOI 복원은 SiPM 피치와 결정 피치가 불일치한 경우에도 균일한 DOI 응답을 보였으며, 광수집 비균일성을 효과적으로 보정하였다. 또한, 제안된 generalized proportional 방법은 ICS 이벤트의 전방·후방 산란 비대칭성과 DOI 의존적 광학 변화를 선형 회귀로 보정함으로써 기존 최대·최소 에너지 및 비례법에 비해 현저히 향상된 정확도를 보였다. Monte Carlo 시뮬레이션에서는 1% 이하의 오차, 실제 실험에서는 2.7–4.5%의 오차를 달성하였다.
이 연구는 multiplexed readout 환경에서도 고정밀 이벤트 위치 복원이 가능함을 실험적으로 입증하였으며, 광전 변환, 신호처리, 통계적 복원의 연계를 통해 PET 검출기의 구조적 효율성과 해상도를 동시에 향상시킬 수 있음을 보여준다. 제안된 방법은 차세대 고해상도 PET 시스템 설계에 적용 가능하며, 저비용·고정밀·시간 동기화형 분자영상 시스템 개발의 기반이 될 것으로 기대된다.

다국어 초록 (Multilingual Abstract)

This dissertation presents a comprehensive study on modeling and correction of event-localization uncertainties in high-resolution PET detectors employing multiplexed side-readout architectures. The research addresses three major aspects: (1) identification and utilization of inter-crystal scatter (ICS) events, (2) depth-of-interaction (DOI) reconstruction in continuous and pixelated scintillator geometries, and (3) optimization of optical confinement and segmentation for enhanced spatial accuracy. A unified analytical and experimental framework was developed to evaluate how optical transport, electronic multiplexing, and statistical modeling jointly determine event-localization precision.
The proposed methods include a multiplexed front-end acquisition circuit, a data-driven DOI reconstruction algorithm based on artificial neural networks (ANN), and a generalized proportional model for ICS positioning. The ANN decoder achieved uniform DOI encoding across mismatched crystal and SiPM pitches, mitigating nonuniform light collection in long-crystal modules. For ICS localization, a regression-based proportional model was formulated to correct forward–backward scattering asymmetry and DOI-dependent optical variation. Monte Carlo simulations and experiments with a 3×3×10 mm³ LGSO side-readout array demonstrated sub-percent positioning errors in simulation and 2.7–4.5% errors in experiment, significantly outperforming conventional maximum-energy, minimum-energy, and proportional methods.
Overall, this work establishes a physically consistent and computationally efficient framework for precise 3D event localization under multiplexed readout. By combining analytical modeling, machine-learning-based reconstruction, and experimental validation, the study demonstrates that high-resolution PET detectors can maintain high accuracy without increasing readout complexity. The proposed methods are expected to contribute to the design of scalable, cost-efficient, and timing-compatible PET systems for next-generation molecular imaging.
ii
Parts of the results in this thesis have been published in the following papers:
Shim, HS et al. "Inter-crystal scattering event identification using a novel silicon photomultiplier signal multiplexing method." Phys. in Med. & Biol. 68.11 (2023): 115008.
Shim, HS, et al. "Continuous single-ended depth-of-interaction measurement using highly multiplexed signals and artificial neural networks." Phys. in Med. & Biol. 70.3 (2025): 035004.

목차 (Table of Contents)

• Abstract i
• Contents iii
• List of Figures vi
• List of Tables x
• Chapter 1. General Introduction 1
• 1.1. Event Localization Uncertainties in PET Detectors 2
• 1.2. Depth-of-Interaction and Parallax Error 3
• 1.3. Inter-Crystal-Scattering and Line-of-Response errors 4
• 1.4. Signal Multiplexing and Machine Learning–Based Decoding 6
• 1.5. Side-Readout Geometry and DOI–ICS Coupling 7
• 1.6. Objectives and Contributions of the Dissertation 8
• Chapter 2. Modeling and Experimental Validation of Inter-Crystal Scattering Identification Using Multiplexed Readout 11
• 2.1. Information Requirements for ICS Analysis 11
• 2.2. Necessity for Information-Preserving Multiplexing 12
• 2.3. High-Pass Filter Multiplexing Architecture 13
• 2.4. Artificial Neural Network for Energy Estimation 15
• 2.5. Experimental Validation and Performance Evaluation 16
• 2.5.1. Experimental Setup 16
• 2.5.2. Energy Resolution and Linearity 17
• 2.5.3. Flood Map and Spatial Resolution Results 18
• 2.5.4. ICS Event Characterization 21
• 2.6. Summary 22
• Chapter 3. Continuous DOI Estimation Using Multiplexed Readout 24
• 3.1. Principle of Continuous DOI Encoding 24
• 3.2. Neural Network-Based DOI Reconstruction 25
• 3.3. Experimental Validation and Performance Evaluation 27
• 3.4. Discussion and Summary 32
• Chapter 4. Influence of Light Confinement and Decoding on Continuous DOI Performance 34
• 4.1. Introduction and Research Scope 34
• 4.2. Experimental Configuration and Optical Parameters 35
• 4.3. Decoding Algorithms and Evaluation Metrics 37
• 4.4. Results 39
• 4.4.1. Scenario 1 – Effect of Array Segmentation 39
• 4.4.1.1. Unsegmented Array 39
• 4.4.1.2. 2 × 2 Segmented Array 41
• 4.4.1.3. 4 × 4 Segmented Array 42
• 4.4.2. Scenario 2 – Effect of Increased Optical Interfaces 43
• 4.4.2.1. 1.5 mm pitch 43
• 4.4.2.2. 3 mm pitch 44
• 4.5. Discussion and Summary 46
• Chapter 5. Generalized ICS Positioning for Side-Readout Geometry Utilizing DOI 49
• 5.1. Background and Motivation 49
• 5.2. Detector Configuration and Multiplexed Readout 50
• 5.3. Methods 53
• 5.3.1. Simulation Study 53
• 5.3.2. Experimental Validation 54
• 5.4. Simulation and Experimental Results 56
• 5.4.1. Simulation Study 57
• 5.4.2. Experimental Validation 61
• 5.5. Discussion and Summary 64
• Chapter 6. General Discussion and Conclusion 67
• Bibliography 71
• Abstract in Korean 76