- 56 views
- 74 downloads
A Comprehensive Study of Conditional Generative Adversarial Networks for Noise Reduction in Optical Coherence Tomography
-
- Author / Creator
- Youwei, Chen
-
Optical coherence tomography (OCT) has been widely adopted as an imaging modality for various clinical applications, such as breast cancer screening, retinal imaging, and vascular assessment, due to its non-invasive nature. However, OCT is affected by coherent speckle noise, which impairs OCT images’ contrast and detailed structural information. This presents a
challenge for accurate clinical analysis. To improve image quality, one can adopt frame-wise averaging of OCT images, where multiple images of the same field of view are consecutively acquired and averaged. However, this approach is time-consuming and impractical for point-of-care clinical applications. To overcome the time-consuming shortcomings and to make OCT more suitable for surgical settings, we propose the application of conditional generative adversarial networks (cGAN). The cGAN was trained to learn how the signal and
noise characteristics change via the averaging process, using non-clinical data for training and then testing on unseen clinical breast tissues. This method has demonstrated strong robustness and generalizability, significantly enhancing signal and contrast without compromising sharpness and reducing speckle noise in OCT B-scans of human breast tissue. The proposed method offers a potential replacement for frame-wise averaging approaches. -
- Graduation date
- Fall 2024
-
- Type of Item
- Thesis
-
- Degree
- Master of Science
-
- License
- This thesis is made available by the University of Alberta Library with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.