Automatic Localization of Brain ROI from MRI using a Two-Stage Ensemble Hough Convolutional Neural Network
- Author(s)
- 아볼 바셔
- Issued Date
- 2020
- Abstract
- Artificial Intelligence is contributing in various ways to resolve complicated problems now a days. Artificial neural network has already shown a lot of break through in different disciplines, such as, medical imaging, computer vision, share market. Deep learning based algorithms are using to address various complex problems in medical imaging. Automatic localization of brain regions of interest (ROIs) from magnetic resonance imaging (MRI) scan is a crucially important task to diagnose the various neuro-degenerative diseases, such as, Alzheimer's disease. In this study, a method has been proposed to automatically locate the brain ROI, such as, hippocampus from MRI scan using a two-stage Hough convolutional neural Network (Hough-CNN). The proposed approach is a amalgamation of Hough voting and the deep convolutional neural network (CNN) and it locates the hippocampus in two Phase. In the first phase, the patches are extracted from the whole MRI scan to train the global Hough-CNN except the boundary region. In the second phase, the local patches are generated in the vicinity of the hippocampus, and then the local models are trained using those patches. In the test phase, the extracted patches from the whole hippocampal region are used to predict the global position of hippocampus. After that, using the global positions of the hippocampus, the local patches are extracted in the vicinity of the target voxel. Using the local patches, the local Hough-CNN estimates the exact location of the hippocampus in the MRI scan. The proposed method is verified using Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Gwangju Alzheimer's and Related Dementia (GARD) cohort datasets.|인공 지능은 복잡한 문제 해결에 다양한 방식으로 기여하고 있습니다. 인공 신경 네트워크는 이미 의료 이미징, 컴퓨터 비전과 같은 다양한 분야에서 많은 돌파구를 보여주었습니다. Magnetic resonance image (MRI) 스캔으로부터 관심있는 뇌 region of interest (ROI)의 자동 위치 예측은 알츠하이머 병과 같은 다양한 신경 퇴행성 질환을 진단하기 위한 기본 작업 중 하나입니다. 이 연구에서는 2 단계 Hough Convolutional Neural Network (Hough-CNN)를 사용하여 MRI 스캔의 해마와 같은 뇌 ROI를 자동으로 찾는 방법이 제안됩니다. 제안 된 접근법은 Hough 투표와 CNN (deep convolutional neural network)의 융합으로 2 단계 Hough-CNN으로 해마를 찾습니다. 첫 번째 단계에서는 전역 Hough-CNN을 훈련시키기 위해 전체 MRI 스캔에서 three-view-patch (TVP)가 추출됩니다. 두 번째 단계에서는 해마 근처에서 로컬 TVP를 생성한 다음 해당 패치를 사용하여 로컬 모델이 학습합니다. 테스트 단계에서, 전체 해마 영역에서 추출 된 TVP는 해마의 전체 위치를 예측하는 데 사용되며, 해마의 전역 위치를 사용하여, 그 주변 로컬 TVP를 추출합니다. 로컬 TVP를 사용하여 로컬 Hough-CNN은 MRI 스캔에서 해마의 정확한 위치를 다시 추정합니다. 제안 된 방법은 ADNI (Alzheimer 's Disease Neuroimaging Initiative) 및 Gwangju Alzheimer 's and Related Dementia (GARD) 코호트 데이터 세트에서 state-of-the-arts 결과를 보여주고 있습니다.
- Alternative Title
- 2 단계 앙상블 허프 컨벌루셔널 뉴럴 네트워크를 사용한 뇌 MRI ROI위치 예측 연구
- Alternative Author(s)
- Abol Basher
- Department
- 일반대학원 컴퓨터공학과
- Advisor
- Professor Jung Ho Yub
- Awarded Date
- 2020-02
- Table Of Contents
- I. INTRODUCTION 1
A. Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
B. The Research Objectives . . . . . . . . . . . . . . . . . . . . . 4
C. Thesis Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
II. Related Works 6
III. Methodology and Datasets 13
A. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
B. Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
C. Patch and Label Generation . . . . . . . . . . . . . . . . . . . . 16
IV. Network Architecture 19
A. Convolutional Neural Network . . . . . . . . . . . . . . . . . . 19
B. Network Architecture Overview . . . . . . . . . . . . . . . . . 21
C. Global Hough Convolutional Neural Network . . . . . . . . . . 22
D. Local Hough Convolutional Neural Network . . . . . . . . . . . 24
E. Loss Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
F. Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
V. Localization Procedures 32
A. ROI Localization Proccedure . . . . . . . . . . . . . . . . . . . 32
B. Error Calculation . . . . . . . . . . . . . . . . . . . . . . . . . 34
C. Statistical Analaysis . . . . . . . . . . . . . . . . . . . . . . . . 35
D. Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
E. Discussion and Comparison . . . . . . . . . . . . . . . . . . . . 49
VI. CONCLUSION 52
PUBLICATIONS 54
A. Journals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
B. Conferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
REFERENCES 64
ACKNOWLEDGEMENTS 65
- Degree
- Master
- Publisher
- Chosun University
- Citation
- 아볼 바셔. (2020). Automatic Localization of Brain ROI from MRI using a Two-Stage Ensemble Hough Convolutional Neural Network.
- Type
- Dissertation
- URI
- https://oak.chosun.ac.kr/handle/2020.oak/14036
http://chosun.dcollection.net/common/orgView/200000278502
-
Appears in Collections:
- General Graduate School > 3. Theses(Master)
- Authorize & License
-
- AuthorizeOpen
- Embargo2020-02-26
- Files in This Item:
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.