KCC 2021
Current Result Document : 2 / 2
ÇѱÛÁ¦¸ñ(Korean Title) |
ÀÇ·á ¿µ»ó¿¡¼ ºñÁ¤»ó Á¶Á÷À» ºÐÇÒÇϱâ À§ÇÑ ¸ÖƼ·¹º§ º¯È¯±â ±â¹Ý ºÐÇÒ ³×Æ®¿öÅ© |
¿µ¹®Á¦¸ñ(English Title) |
Multi-Level Transformer-based Segmentation Network for Abnormal Tissue Segmentation in Medical Image |
ÀúÀÚ(Author) |
µÎÀÌ Çª¿Ë ´Ù¿À
¾çÇüÁ¤
±è¼öÇü±¹¹®¸í
À̱ͻó
°¼¼·É
¿ÀÀÎÀç
Duy-Phuong Dao
Hyung-Jeong Yang
Soo-Hyung Kim
Guee-Sang Lee
Sae-Ryung Kang
In-jae Oh
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 48 NO. 01 PP. 0700 ~ 0702 (2021. 06) |
Çѱ۳»¿ë (Korean Abstract) |
|
¿µ¹®³»¿ë (English Abstract) |
In medical image, abnormal tissue segmentation is an essential technique for developing healthcare systems, particularly for disease diagnosis and treatment planning. Recently, inspired by advance in deep learning, many architecture networks, designed for segmentation tasks, were introduced and achieved remarkable performance. However, most of those methods are based on convolution operation which only has an ability to capture locality features. In this paper, we propose a Transformer-based segmentation network which can not only capture global context but also integrate multi-level features at encoder branch. The experimental results prove our proposal outperform compared to conventional methods in terms of Dice score. |
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|