.png){kind=icon}
.png){kind=icon}
Return
ECLNet: Efficient convolution with lite transformer for thymoma segmentation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.displa.2025.103091.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• We propose ECLNet, a hybrid architecture combining efficient convolution and lightweight transformer to reduce computational overhead. • We propose lightweight global self-attention to capture global dependencies in high-resolution medical images while reducing computation. • We propose a refined block in the output layer to adjust depth features, enhancing tumor boundary accuracy and thymoma segmentation precision. • We propose using ECLNet for thymoma segmentation, achieving state-of-the-art performance validated on BTCV and DBCM datasets.