Skip to main content
Fig. 1 | Radiation Oncology

Fig. 1

From: Pelvic U-Net: multi-label semantic segmentation of pelvic organs at risk for radiation therapy anal cancer patients using a deeply supervised shuffle attention convolutional neural network

Fig. 1

Proposed, deeply supervised Pelvic U-Net architecture for organs at risk (OAR) segmentation in the pelvic region (a). 3D patches with a size of \(1 \times 80 \times 160 \times 160\) pixels are extracted from computed tomography (CT) image volumes and used as the encoder input. A series of convolutional and max pooling operations is then applied to the input patch for feature extraction purposes. Feature map upscaling in the decoder part is performed using trilinear interpolation. High level features from the encoder are copied and concatenated with low level features using skip connections. In addition, shuffle attention (SA) blocks are incorporated into the skip connections, combining spatial and channel attention (b)

Back to article page

Radiation Oncology

ISSN: 1748-717X

Contact us