Diffusion and Multi-Domain Adaptation Methods for Eosinophil Segmentation
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909139560562688 |
|---|---|
| author | Lin, Kevin Brown, Donald Syed, Sana Greene, Adam |
| author_facet | Lin, Kevin Brown, Donald Syed, Sana Greene, Adam |
| contents | Eosinophilic Esophagitis (EoE) represents a challenging condition for medical providers today. The cause is currently unknown, the impact on a patient's daily life is significant, and it is increasing in prevalence. Traditional approaches for medical image diagnosis such as standard deep learning algorithms are limited by the relatively small amount of data and difficulty in generalization. As a response, two methods have arisen that seem to perform well: Diffusion and Multi-Domain methods with current research efforts favoring diffusion methods. For the EoE dataset, we discovered that a Multi-Domain Adversarial Network outperformed a Diffusion based method with a FID of 42.56 compared to 50.65. Future work with diffusion methods should include a comparison with Multi-Domain adaptation methods to ensure that the best performance is achieved. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_11323 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Diffusion and Multi-Domain Adaptation Methods for Eosinophil Segmentation Lin, Kevin Brown, Donald Syed, Sana Greene, Adam Image and Video Processing I.4.6 Eosinophilic Esophagitis (EoE) represents a challenging condition for medical providers today. The cause is currently unknown, the impact on a patient's daily life is significant, and it is increasing in prevalence. Traditional approaches for medical image diagnosis such as standard deep learning algorithms are limited by the relatively small amount of data and difficulty in generalization. As a response, two methods have arisen that seem to perform well: Diffusion and Multi-Domain methods with current research efforts favoring diffusion methods. For the EoE dataset, we discovered that a Multi-Domain Adversarial Network outperformed a Diffusion based method with a FID of 42.56 compared to 50.65. Future work with diffusion methods should include a comparison with Multi-Domain adaptation methods to ensure that the best performance is achieved. |
| title | Diffusion and Multi-Domain Adaptation Methods for Eosinophil Segmentation |
| topic | Image and Video Processing I.4.6 |
| url | https://arxiv.org/abs/2403.11323 |