Cited time in webofscience Cited time in scopus

Full metadata record

DC Field Value Language
dc.contributor.advisor 문인규 -
dc.contributor.author Sungwoo Son -
dc.date.accessioned 2022-03-07T16:00:04Z -
dc.date.available 2022-03-07T16:00:04Z -
dc.date.issued 2022 -
dc.identifier.uri http://dgist.dcollection.net/common/orgView/200000595277 en_US
dc.identifier.uri http://hdl.handle.net/20.500.11750/16259 -
dc.description Cancer cell classification, Holographic image analysis, Deep learning, Convolutional neural network -
dc.description.abstract Diagnosing cancer is one of the most important topics in medical field. There are lots of conventional methods of diagnosing, but cancer cell classification is very challenging due to morphological similarity. Digital holography in a microscopic configuration can provide quantitative phase image representing the intracellular content and morphology of cells. In this paper, I suggest a method of classification of three types of cancer cell lines (lung, breast, and skin) by image-based deep learning with a convolutional neural network (CNN) and digital holography. I trained two type of deep learning CNN model (without the skip connection) and Resnet (with skip connection). I compared image-based classification result with feature-based classification (random forest, support vector machine and pattern recognition artificial neural networks). And I compared fluorescent image (internal cell morphology) classification for further applicability. I analyzed phase image-based classification model outperformed feature-based classification by about 9% and fluorescent image classification by about 12% higher accuracy. I expect by using image-based classification model cancer can be diagnosed faster and more accurately.|암 진단은 의학 분야에서 가장 중요한 주제 중 하나이다. 기존 진단 방법은 많이 있지만 형태학적 유사성으로 인해 암세포 분류가 매우 어렵다. 현미경 구성의 디지털 홀로그래피는 세포 내 함량 및 세포 형태를 나타내는 정량적 위상 이미지를 제공할 수 있다. 본 논문에서는 CNN (Convolutional Neural Network) 과 디지털 홀로그래피를 이용한 이미지 기반 딥러닝을 통해 3가지 유형의 암세포 (폐, 유방, 피부) 를 분류하는 방법을 제안한다. 두 가지 유형의 딥러닝 CNN 모델 (skip connction 없음) 과 Resnet (skin connection 있음) 을 훈련하였다. 이미지 기반 딥러닝 기법 분류 결과를 특징 기반 분류 기법들 (랜덤 포레스트, 지원 벡터 머신 및 패턴 인식 인공 신경망) 과 비교하였다. 그리고 추가적용을 위해 형광영상 (내부 세포 형태) 분류를 비교하였다. 위상 이미지 기반 분류 모델은 특징 기반 분류보다 약 9%, 형광 이미지 분류보다 약 12% 더 높은 정확도로 분석하였다. 이미지 기반 분류 모델을 사용하여 암을 더 빠르고 정확하게 진단할 수 있을 것으로 기대한다. -
dc.description.statementofresponsibility Y -
dc.description.tableofcontents Ⅰ. INTRODUCTION 1
Ⅱ. DIGITAL HOLOGRAPHIC MICROSCOPY(DHM) 4
Ⅲ. SAMPLE PREPARATION AND DATA GENERATION 7
3.1 Cancer cell sample preparation 7
3.2 Fluorescent image generation 9
Ⅳ. METHOD 9
4.1 Feature-based classification 9
4.1.1 Random forest classifier 9
4.1.2 Support vector machine 10
4.1.3 Pattern recognition artificial neural network(PR-ANN) 10
4.2 Image-based classification 12
4.2.1 Convolutional neural network(CNN) 12
4.2.2 Resnet 14
Ⅴ. EXPERIMENT RESULTS 15
5.1 Feature-based classification 15
5.2 Image-based classification 19
Ⅵ. CONCLUSION 21
References 23
-
dc.format.extent 27 -
dc.language eng -
dc.publisher DGIST -
dc.source /home/dspace/dspace53/upload/200000595277.pdf -
dc.subject Cancer cell classification, Holographic image analysis, Deep learning, Convolutional neural network -
dc.title Automated image-based classification of cancer cell by using digital holography and deep learning -
dc.title.alternative 디지털 홀로그래피와 딥러닝을 이용한 이미지 기반 암세포 자동화 분류 -
dc.type Thesis -
dc.identifier.doi 10.22677/thesis.200000595277 -
dc.description.degree Master -
dc.contributor.department Robotics Engineering -
dc.contributor.coadvisor Okkyun Lee -
dc.date.awarded 2022/02 -
dc.publisher.location Daegu -
dc.description.database dCollection -
dc.citation XT.RM 손57 202202 -
dc.date.accepted 1/21/22 -
dc.contributor.alternativeDepartment 로봇공학전공 -
dc.embargo.liftdate 20220117 -
dc.contributor.affiliatedAuthor Sungwoo Son -
dc.contributor.affiliatedAuthor Inkyu Moon -
dc.contributor.affiliatedAuthor Okkyun Lee -
dc.contributor.alternativeName 손승우 -
dc.contributor.alternativeName Inkyu Moon -
dc.contributor.alternativeName 이옥균 -
Files in This Item:

There are no files associated with this item.

Appears in Collections:
Department of Robotics and Mechatronics Engineering Theses Master

qrcode

  • twitter
  • facebook
  • mendeley

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

BROWSE