Tumor attention networks: Better feature selection, better tumor segmentation

Shuchao Pang, Anan Du, Mehmet A. Orgun, Yunyun Wang, Zhenmei Yu

Abstract: Highlights•We propose an accurate automatic tumor segmentation approach called TA-NET.•This work fully leverages the current popular Convolutional Neural Networks (CNNs).•TA-Net features a tumor attention layer to imitate the human attention mechanism.•TA-Net can identify valuable features and suppress unrelated ones in liver scans.•This work conducts comprehensive experiments on three real-world medical datasets.