Title

Optimized Red Blood Cell Segmentation in Holographic Imaging through Integration of Self-Supervised Learning and Diffusion Models

DGIST Authors

Hyunbin An ; Inkyu Moon ; Youhyun Kim

Advisor

문인규

Co-Advisor(s)

Youhyun Kim

Issued Date

2025

Awarded Date

2025-02-01

Citation

Hyunbin An. (2025). Optimized Red Blood Cell Segmentation in Holographic Imaging through Integration of Self-Supervised Learning and Diffusion Models. doi: 10.22677/THESIS.200000828659

Type

Thesis

Description

Red blood cells, Digital Holography, Deep Learning, Diffusion Model, Self-Supervised learning

Table Of Contents

I. INTRODUCTION 1
II. METHODOLOGY 6
2.1 Data Acquisition 6
2.2 Digital Holographic Microscopy 7
2.3 Synthetic RBC Image Generation Using Diffusion Model 9
2.4 Self-Supervised Learning for Pretrained Model 13
2.5 Watershed Algorithm 16
2.6 Evaluation Metrics 17
III. EXPERIMENTAL SETUPS 18
3.1 Datasets 18
3.2 Implementation Setups 18
IV. RESULTS 20
4.1 Generated Synthetic RBC Data 20
4.2 Semantic Segmentation 21
4.3 Effectiveness of Losses in Self-Supervised Learning 24
4.4 Experiments on the Effect of SSCRL in Limited Training Data 25
4.5 Phenotypical Assessment of Red Blood Cells 26
V. CONCLUSION 27

URI

http://hdl.handle.net/20.500.11750/58017
http://dgist.dcollection.net/common/orgView/200000828659

DOI

10.22677/THESIS.200000828659

Degree

Master

Department

Department of Robotics and Mechatronics Engineering

Publisher

DGIST

Show Full Item Record