View : 85 Download: 0

Full metadata record

DC Field Value Language
dc.contributor.advisor민동보-
dc.contributor.author윤지원-
dc.creator윤지원-
dc.date.accessioned2026-09-02T16:31:51Z-
dc.date.available2026-09-02T16:31:51Z-
dc.date.issued2026-
dc.identifier.otherOAK-000000262273-
dc.identifier.urihttps://dcollection.ewha.ac.kr/common/orgView/000000262273en_US
dc.identifier.urihttps://dspace.ewha.ac.kr/handle/2015.oak/276251-
dc.description.abstractAccurate focus quality assessment (FQA) in fluorescence microscopy is challenging due to stain-dependent optical variations that induce heterogeneous focus behavior across images. Existing methods, however, treat focus quality as a stain-agnostic problem, assuming a shared global ordering. We formulate stain-aware FQA for fluorescence microscopy, showing that focus-rank relationships vary substantially across stains due to stain-dependent imaging characteristics and invalidate this assumption. To support this formulation, we introduce FluoMix, the first dataset for stain-aware FQA spanning multiple tissues, fluorescent stains, and focus levels. We further propose FluoCLIP, a two-stage vision-language framework that grounds stain semantics and enables stain-conditioned ordinal reasoning for focus prediction, effectively decoupling stain representation from ordinal structure. By explicitly modeling stain-dependent focus behavior, FluoCLIP consistently outperforms both conventional FQA methods and recent vision-language baselines, demonstrating strong generalization across diverse fluorescence microscopy conditions. Code and dataset are publicly available at https://fluoclip.github.io/.;형광 현미경에서 정확한 초점 품질 평가(Focus Quality Assessment, FQA)는 염색(Stain) 의존적 광학 변이로 인해 이미지마다 서로 다른 초점 거동이 유발되므로 매우 까다로운 과제이다. 그러나 기존 방법들은 초점 품질을 염색(Stain)과 무관한 문제로 취급하며, 모든 이미지가 공유된 전역 순서(global ordering)를 따른다고 가정한다. 본 연구에서는 형광 현미경을 위한 염색 기반 초점평가(Stain-Aware FQA)를 새롭게 정의하고, 염색(Stain)별 영상(Imaging)의 특성에 따라 초점-순위 관계(focus-rank relationship)가 크게 달라진다는 점을 밝혀 기존의 가정이 타당하지 않음을 입증한다. 이러한 정식화를 뒷받침하기 위해, 다양한 조직, 형광 염색(Stain) 및 초점 단계를 아우르는 염색 기반 초점평가(Stain-Aware FQA)를 위한 최초의 데이터셋인 FluoMix를 소개한다. 나아가, 본 연구는 염색(Stain) 의미론을 기반으로 하여 초점 예측을 위한 염색(Stain) 조건부 서수 추론(ordinal reasoning)을 수행하는 2단계 시각-언어(vision-language) 프레임워크인 FluoCLIP을 제안한다. 이 모델은 서수(ordinal) 구조로부터 염색(Stain) 표현을 효과적으로 분리해낸다. 염색(Stain) 의존적 초점 거동(stain-dependent focus behavior)을 명시적으로 모델링함으로써, FluoCLIP은 기존의 초점 평가 방법론과 최신 시각-언어 베이스라인 모델들을 일관되게 능가하는 성능을 보였으며, 다양한 형광 현미경 조건에서도 강력한 일반화 성능을 입증하였다. 관련 코드와 데이터셋은 다음 링크에서 확인할 수 있다: https://fluoclip.github.io/.-
dc.description.tableofcontentsⅠ. Introduction 1 Ⅱ. Related Work 4 1. Microscopy Focus Quality Assessment 4 2. CLIP-based Ordinal Regression 4 3. Biomedical Vision-Language Models 5 Ⅲ. Methods 6 1. Problem Statement 6 2. FluoCLIP: Stain-Aware FQA Framework 7 1. Stage 1: Stain-Grounding 8 2. Stage 2: Stain-Guided Ranking 9 3. Overall Training Objective 10 4. Algorithm 12 Ⅳ. FluoMix for Stain-Aware FQA 13 1. Motivation and Design Principles 13 2. FluoMix Composition 13 3. Annotation Protocol 14 4. Biological Preparation and Imaging Protocols 15 1. Tissue Preparation 15 2. Immunostaining 15 3. Imaging Conditions 15 Ⅴ. Empirical Evidence for Stain-Dependent Focus Behavior 17 1. Per-Rank Behavior of Sharpness Metric 17 2. Stain-Wise Variability Across Datasets 18 3. Sharpness Metric and Rank Correlation 18 Ⅵ. Experiments 20 1. Datasets 20 1. BBBC006 20 2. FocusPath 20 3. FluoMix 20 2. Experiment Settings 21 1. Training Details 21 2. Baselines 23 3. Evaluation Metrics 23 3. Comparative Study 25 1. Results on FluoMix 25 2. Results on BBBC006 25 3. Results on FocusPath 25 4. Ablation Study 26 1. Contribution of Individual Components 27 2. Effects of Stain-Grounding 27 3. Effects of Adapter 27 4. Combined Effects 28 5. Effects of Learning Stain Representations 28 5. Few-Shot Validations 29 1. Few-Shot Learning on FluoMix 29 2. Few-Shot Generalization to Unseen Tissues 31 3. Few-Shot Generalization to Unseen Stain & Tissue Pairs 32 Ⅶ. Conclusion 33 Ⅷ. Future Work 34 Bibliography 36-
dc.formatapplication/pdf-
dc.format.extent6381082 bytes-
dc.languageeng-
dc.publisher이화여자대학교 대학원-
dc.subjectFocus Quality Assessment, Biomedical Imaging, Multimodal Learning-
dc.subject.ddc000-
dc.titleFluoCLIP-
dc.typeMaster's Thesis-
dc.title.subtitleStain-Aware Focus Quality Assessment in Fluorescence Microscopy-
dc.creator.othernameYoon, Jiwon-
dc.format.pagevi, 39 p.-
dc.identifier.thesisdegreeMaster-
dc.identifier.major대학원 인공지능·소프트웨어학부-
dc.date.awarded2026. 8-
Appears in Collections:
ETC > ETC
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

BROWSE