PointRendUNet: a model for fine boundary segmentation

Linlin Li; Xinzhuo Zhao; Li Ke

doi:10.1117/12.3009464

11 October 2023 PointRendUNet: a model for fine boundary segmentation

Linlin Li, Xinzhuo Zhao, Li Ke

Proceedings Volume 12918, Fourth International Conference on Computer Science and Communication Technology (ICCSCT 2023); 129181Q (2023) https://doi.org/10.1117/12.3009464
Event: International Conference on Computer Science and Communication Technology (ICCSCT 2023), 2023, Wuhan, China

Abstract

Segmenting tumors from brain MRI sequences is critical for prognosis assessment, radiomics analysis, surgical planning, and pathological diagnosis. However, semantic segmentation based on convolutional neural networks is on the regular grid representation of the input and output of the network, which leads to the tendency of the predicted label edges to smooth. The infiltrating nature of malignant gliomas also poses challenges for the segmentation task. Therefore, we propose an automated, standardized method for fine-segmenting brain tumor boundaries. This method introduces the PointRend module based on 3D-UNet and trains a model with coarse-to-fine segmentation on the open dataset BraTS. The performance of our model on the public validation dataset is as follows: Dice of 0.84699, 0.84640, and 0.84661 for the whole tumor, tumor core, and enhancing tumor area, respectively. The evaluation shows that our model produces a good and balanced performance for different tumor subregions.

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Linlin Li, Xinzhuo Zhao, and Li Ke "PointRendUNet: a model for fine boundary segmentation", Proc. SPIE 12918, Fourth International Conference on Computer Science and Communication Technology (ICCSCT 2023), 129181Q (11 October 2023); https://doi.org/10.1117/12.3009464

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
6 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Tumors

Image segmentation

Education and training

3D modeling

Data modeling

Image quality

Semantics

Show All Keywords

Keywords/Phrases

Search In:

Publication Years