Abstract—In this study, we propose an automated method for assessing small bowel motility function with cine MRI using 3D U-Net, which is a kind of deep fully convolutional neural networks for 3D semantic segmentation. In the proposed method, the cine MR images (temporal MR image sequence) is treated as a 3D image. We applied 3D U-Net, which employs 3D convolution, to automatically segment the temporal small bowel image sequence. Compared with the conventional 2D U-Net, in which the small bowel was segmented without temporal information and just segmented frame by frame, the proposed 3D U-Net can accurately and simultaneously segment all frames using temporal information. This is the first 3D fully convolutional network for small bowel segmentation in cine MR images (temporal sequence images), to the best of our knowledge. The small bowel motility function is assessed by the use of the segmented temporal MR image sequence. Experimental results demonstrate the effectiveness of the proposed method.
Index Terms—cine-MR image, fully convolutional neural network, U-Net, 3D, small bowel, motility
Cite: Kazuki Otsuki, Yutaro Iwamoto, Yen-Wei Chen, Akira Furukawa, and Shuzo Kanasaki, "Cine-MR Image Segmentation for Assessment of Small Bowel Motility Function Using 3D U-Net," Journal of Image and Graphics, Vol. 7, No. 4, pp. 134-139, December 2019. doi: 10.18178/joig.7.4.134-139
Copyright © 2012-2020 Journal of Image and Graphics, All Rights Reserved