BFT‐Net: A transformer‐based boundary feedback network for kidney tumour segmentation DOI Creative Commons
Tianyu Zheng, Chao Xu, Zhengping Li

et al.

IET Communications, Journal Year: 2024, Volume and Issue: 18(16), P. 966 - 977

Published: July 12, 2024

Abstract Kidney tumours are among the top ten most common tumours, automatic segmentation of medical images can help locate tumour locations. However, kidney still faces several challenges: firstly, there is a lack renal endoscopic datasets and no techniques for images; secondly, intra‐class inconsistency caused by variations in size, location, shape tumours; thirdly, difficulty semantic fusion during decoding; finally, issue boundary blurring localization lesions. To address aforementioned issues, new dataset called Re‐TMRS proposed, this dataset, transformer‐based feedback network (BFT‐Net) proposed. This incorporates an adaptive context extract module (ACE) to emphasize local contextual information, reduces gap through mixed feature capture (MFC), ultimately improves extraction capability end‐to‐end optimization learning assist (BA). Through numerous experiments, it demonstrated that proposed model exhibits excellent ability generalization performance. The mDice mIoU on reach 91.1% 91.8%, respectively.

Language: Английский

BFT‐Net: A transformer‐based boundary feedback network for kidney tumour segmentation DOI Creative Commons
Tianyu Zheng, Chao Xu, Zhengping Li

et al.

IET Communications, Journal Year: 2024, Volume and Issue: 18(16), P. 966 - 977

Published: July 12, 2024

Abstract Kidney tumours are among the top ten most common tumours, automatic segmentation of medical images can help locate tumour locations. However, kidney still faces several challenges: firstly, there is a lack renal endoscopic datasets and no techniques for images; secondly, intra‐class inconsistency caused by variations in size, location, shape tumours; thirdly, difficulty semantic fusion during decoding; finally, issue boundary blurring localization lesions. To address aforementioned issues, new dataset called Re‐TMRS proposed, this dataset, transformer‐based feedback network (BFT‐Net) proposed. This incorporates an adaptive context extract module (ACE) to emphasize local contextual information, reduces gap through mixed feature capture (MFC), ultimately improves extraction capability end‐to‐end optimization learning assist (BA). Through numerous experiments, it demonstrated that proposed model exhibits excellent ability generalization performance. The mDice mIoU on reach 91.1% 91.8%, respectively.

Language: Английский

Citations

0