Data-driven approach for task-driven medical image reconstruction and analysis

Deep learning is an important part of artificial intelligence, where the neural network can be an efficient way to solve complex problems, especially in the field of computer vision. Therefore, it is benefited to be applied in medical image processing. In this report, we mainly talk about deep MRI r...

Full description

Saved in:
Bibliographic Details
Main Author: Li, Changhao
Other Authors: Wen Bihan
Format: Thesis-Master by Coursework
Language:English
Published: Nanyang Technological University 2020
Subjects:
Online Access:https://hdl.handle.net/10356/141141
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-141141
record_format dspace
spelling sg-ntu-dr.10356-1411412023-07-04T16:42:13Z Data-driven approach for task-driven medical image reconstruction and analysis Li, Changhao Wen Bihan School of Electrical and Electronic Engineering bihan.wen@ntu.edu.sg Engineering::Electrical and electronic engineering Deep learning is an important part of artificial intelligence, where the neural network can be an efficient way to solve complex problems, especially in the field of computer vision. Therefore, it is benefited to be applied in medical image processing. In this report, we mainly talk about deep MRI reconstruction. Base on the compressed sensing magnetic resonance imaging and data-driven method, it benefits patients a lot since the time of MRI acquisition can be largely reduced which only demands k-spaced data with low under sampled rate. Recently, deep learning based method on MRI reconstruction become more popular than traditional approaches. However, most of reconstruction models may not have good generalization performance, which means only perform well on a specific dataset. Thus, it is necessary to figure out how the different deep learning method work and also important to apply them on different dataset. In this report, we introduce the basic principle of deep learning, neural network and MRI reconstruction. Then we present how to apply them on MRI reconstruction. Finally, we would conduct some popular MRI reconstruction models which based on fully convolutional network or generative adversarial network (GAN) or cascaded network, and test atypical network on MRBrainS13 dataset which is not the original dataset of all the referred model in this report. Master of Science (Communications Engineering) 2020-06-04T06:34:35Z 2020-06-04T06:34:35Z 2020 Thesis-Master by Coursework https://hdl.handle.net/10356/141141 en ISM-DISS-01858 application/pdf Nanyang Technological University
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic Engineering::Electrical and electronic engineering
spellingShingle Engineering::Electrical and electronic engineering
Li, Changhao
Data-driven approach for task-driven medical image reconstruction and analysis
description Deep learning is an important part of artificial intelligence, where the neural network can be an efficient way to solve complex problems, especially in the field of computer vision. Therefore, it is benefited to be applied in medical image processing. In this report, we mainly talk about deep MRI reconstruction. Base on the compressed sensing magnetic resonance imaging and data-driven method, it benefits patients a lot since the time of MRI acquisition can be largely reduced which only demands k-spaced data with low under sampled rate. Recently, deep learning based method on MRI reconstruction become more popular than traditional approaches. However, most of reconstruction models may not have good generalization performance, which means only perform well on a specific dataset. Thus, it is necessary to figure out how the different deep learning method work and also important to apply them on different dataset. In this report, we introduce the basic principle of deep learning, neural network and MRI reconstruction. Then we present how to apply them on MRI reconstruction. Finally, we would conduct some popular MRI reconstruction models which based on fully convolutional network or generative adversarial network (GAN) or cascaded network, and test atypical network on MRBrainS13 dataset which is not the original dataset of all the referred model in this report.
author2 Wen Bihan
author_facet Wen Bihan
Li, Changhao
format Thesis-Master by Coursework
author Li, Changhao
author_sort Li, Changhao
title Data-driven approach for task-driven medical image reconstruction and analysis
title_short Data-driven approach for task-driven medical image reconstruction and analysis
title_full Data-driven approach for task-driven medical image reconstruction and analysis
title_fullStr Data-driven approach for task-driven medical image reconstruction and analysis
title_full_unstemmed Data-driven approach for task-driven medical image reconstruction and analysis
title_sort data-driven approach for task-driven medical image reconstruction and analysis
publisher Nanyang Technological University
publishDate 2020
url https://hdl.handle.net/10356/141141
_version_ 1772829099975245824