Fig. 2From: Automatic segmentation of fat metaplasia on sacroiliac joint MRI using deep learningFramework of the deep-learning-based segmentation and classification models. The 2.5D-AttentionUnet segmentation model was developed by combining two- and three-dimensional convolution. The automatic classification model was developed with the assistance of the segmentation results. axSpAÂ axial spondyloarthritis; non-axSpAÂ non-axial spondyloarthritis; ROIÂ region of interestBack to article page