Medical image synthesis with advanced deep learning methods
Synthetic images can be used to assist many applications, such as helping improve the performance of classifiers, generation of realistic images to augment the data, which are difficult in some cases for medical image annotation and acquisition. This report studies the synthesis of heart computed t...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project (FYP) |
| Language: | English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/77251 |
| _version_ | 1811688420601233408 |
|---|---|
| author | Zhang, Zeng |
| author2 | Lin Zhiping |
| author_facet | Lin Zhiping Zhang, Zeng |
| author_sort | Zhang, Zeng |
| collection | NTU |
| description | Synthetic images can be used to assist many applications, such as helping improve the performance of classifiers, generation of realistic images to augment the
data, which are difficult in some cases for medical image annotation and acquisition. This report studies the synthesis of heart computed tomography coronary angiography (CTCA) images with coronary arteries using R-sGAN network [1].
With a generator and discriminator, the R-sGAN network can generate images
with the same content as the query image and the similar style as the style image.
Each 3D CT image in the training and style datasets are split into 2D slices. The
generator creates the synthesized image from random noise, and the discriminator is trained to tell the input being synthetic or real. We have experiments
on the CTCA images. Illustrated by the results on the center part of CTCA, we
show that the synthetic images are similar to the style CT data, and the coronary
content can also be embedded into the synthetic images. A future application is
on tiny structure segmentation such as coronary artery. Because the structure is
very small and yet long across 3D space, annotation becomes a challenge. The
synthetic image might give us a new way to increase the annotated dataset. |
| first_indexed | 2024-10-01T05:31:56Z |
| format | Final Year Project (FYP) |
| id | ntu-10356/77251 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T05:31:56Z |
| publishDate | 2019 |
| record_format | dspace |
| spelling | ntu-10356/772512023-07-07T16:10:50Z Medical image synthesis with advanced deep learning methods Zhang, Zeng Lin Zhiping School of Electrical and Electronic Engineering Institute for Infocomm Research DRNTU::Engineering::Electrical and electronic engineering Synthetic images can be used to assist many applications, such as helping improve the performance of classifiers, generation of realistic images to augment the data, which are difficult in some cases for medical image annotation and acquisition. This report studies the synthesis of heart computed tomography coronary angiography (CTCA) images with coronary arteries using R-sGAN network [1]. With a generator and discriminator, the R-sGAN network can generate images with the same content as the query image and the similar style as the style image. Each 3D CT image in the training and style datasets are split into 2D slices. The generator creates the synthesized image from random noise, and the discriminator is trained to tell the input being synthetic or real. We have experiments on the CTCA images. Illustrated by the results on the center part of CTCA, we show that the synthetic images are similar to the style CT data, and the coronary content can also be embedded into the synthetic images. A future application is on tiny structure segmentation such as coronary artery. Because the structure is very small and yet long across 3D space, annotation becomes a challenge. The synthetic image might give us a new way to increase the annotated dataset. Bachelor of Engineering (Electrical and Electronic Engineering) 2019-05-22T04:59:51Z 2019-05-22T04:59:51Z 2019 Final Year Project (FYP) http://hdl.handle.net/10356/77251 en Nanyang Technological University 64 p. application/pdf |
| spellingShingle | DRNTU::Engineering::Electrical and electronic engineering Zhang, Zeng Medical image synthesis with advanced deep learning methods |
| title | Medical image synthesis with advanced deep learning methods |
| title_full | Medical image synthesis with advanced deep learning methods |
| title_fullStr | Medical image synthesis with advanced deep learning methods |
| title_full_unstemmed | Medical image synthesis with advanced deep learning methods |
| title_short | Medical image synthesis with advanced deep learning methods |
| title_sort | medical image synthesis with advanced deep learning methods |
| topic | DRNTU::Engineering::Electrical and electronic engineering |
| url | http://hdl.handle.net/10356/77251 |
| work_keys_str_mv | AT zhangzeng medicalimagesynthesiswithadvanceddeeplearningmethods |