.png){kind=icon}
.png){kind=icon}
Return
A Transfer Learning-Based Approach for Pheochromocytoma and Surrounding Multi-Organ Segmentation in Abdominal CT Images
.png){kind=icon}
PRE
AI
G
J
E
B
X
J
M
C
DOI：10.1016/j.knosys.2026.116841.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• We integrated a convolutional branch into the generalist model to output detail-supplementing features, while incorporating multi-level spatial adapters in the main Vision Transformer branch to learn task-specific inductive biases and generate spatially-informed features. These dual-branch outputs are then processed through cross-attention mechanisms, significantly enhancing the model's capability to identify crucial local features. • By extracting multi-scale features from intermediate layers of the image encoder and fusing them with decoder-generated masks, we achieved global-local feature integration, thereby optimizing segmentation quality for pheochromocytomas and surrounding organs. • We implemented a transfer learning approach to develop an end-to-end model for precise segmentation of pheochromocytomas and adjacent multi-organ structures. • Experimental validation demonstrates that our single-model solution effectively improves segmentation accuracy for multiple targets in medical images, including tumors, multi-organ structures, and renal vessels.
Keywords:
Medical image segmentation
Transfer learning
Pheochromocytoma
Abdominal multi-organ
End-to-End model
Journal
K
IF:
7.6
Papers:
1.2W
Citations:
4.5W
Organization
Cited Papers