Multi-Scale Feature Aggregation Attention for Pancreatic Segmentation

Accurate pancreatic segmentation of CT images is very important for the observation and intervention of pancreatic cancer. However, due to the complex location and small volume of the pancreas, this task presents significant challenges. To address these issues, we propose a novel multi-scale feature aggregation attention (MFAA) network, which integrates features from different kernels to enhance the performance of pancreatic segmentation. We evaluated our network on the NIH Pancreas dataset, and experimental results show that MFAA outperforms existing methods in multiple evaluation metrics and performs well across four metrics.