Skip to main content
Fig. 5 | Radiation Oncology

Fig. 5

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. 5

Observations made during a more detailed, visual comparison, between the segmentation structures predicted by the Pelvic U-Net and the manual, ground truth (GT), delineations for multiple test patients. In some 2D image slices, tissue near the actual bone structures was not segmented (a–c). Furthermore, incorrect delineations caused by the automatic, thresholding-based segmentation algorithm used for generating the GT bone marrow data were found (d, e). In addition, rare, manual delineation errors for e.g., the bladder could be observed (f, g)

Back to article page

Radiation Oncology

ISSN: 1748-717X

Contact us