Dynamic U-Net for multi-organ nucleus segmentation.
Saved in:
| Title: | Dynamic U-Net for multi-organ nucleus segmentation. |
|---|---|
| Authors: | Traisuwan, Attasuntorn1 (AUTHOR) attasuntorn@gmail.com, Limsiroratana, Somchai1 (AUTHOR) somchai.l@psu.ac.th, Phukpattaranont, Pornchai2 (AUTHOR) pornchai.p@psu.ac.th, Tandayya, Pichaya1 (AUTHOR) pichaya.t@psu.ac.th |
| Source: | Multimedia Tools & Applications. Aug2025, Vol. 84 Issue 26, p31537-31564. 28p. |
| Subjects: | Cell segmentation, Image segmentation, Cancer diagnosis, Convolutional neural networks |
| Abstract: | Multi-organ nucleus segmentation is an important task for diagnosis, staging, and grading cancer. It is the first step of many quantitative data analysis pipelines. Most existing methods did not apply frequency-related features to extract the key spatial information. In this work, we propose the Dynamic U-Net, a novel method for multi-organ nucleus segmentation. Specifically, we design the Dynamic Multi-spectral module which is a combination of the Dynamic Convolution and the Multi-spectral Channel Attention module. The Dynamic Convolution enhances the representation capability by aggregating multiple parallel convolution kernels dynamically based upon their attentions. The attention applied inside was the Multi-spectral modules. It maintains useful features in the network. Experiments on MoNuSeg datasets show that our Dynamic U-Net performed slightly better than state-of-the-art of U-Nets. In contrast, the network size is far less costly. [ABSTRACT FROM AUTHOR] |
| Copyright of Multimedia Tools & Applications is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Multi-organ nucleus segmentation is an important task for diagnosis, staging, and grading cancer. It is the first step of many quantitative data analysis pipelines. Most existing methods did not apply frequency-related features to extract the key spatial information. In this work, we propose the Dynamic U-Net, a novel method for multi-organ nucleus segmentation. Specifically, we design the Dynamic Multi-spectral module which is a combination of the Dynamic Convolution and the Multi-spectral Channel Attention module. The Dynamic Convolution enhances the representation capability by aggregating multiple parallel convolution kernels dynamically based upon their attentions. The attention applied inside was the Multi-spectral modules. It maintains useful features in the network. Experiments on MoNuSeg datasets show that our Dynamic U-Net performed slightly better than state-of-the-art of U-Nets. In contrast, the network size is far less costly. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 13807501 |
| DOI: | 10.1007/s11042-024-20444-z |
