Renal Parenchyma Segmentation in Abdominal MR Images based on Cascaded Deep Neural Networks with Image and Shape Patches

Hyeonjin Kim, Helen Hong, Dae Chul Jung, Kidon Chang, Koon Ho Rha

Keywords: Magnetic resonance (MR), Renal parenchyma, Deep neural network, Image patch, Shape patch

Abstract: We propose an automatic segmentation method of renal parenchyma in abdominal MR images based on cascaded deep neural networks with image and shape patches. First, intensity and spacing normalization is performed in the whole MR image. Second, kidney is localized with the ensemble of 2D segmentation networks based on attention U-Net on the axial, coronal, sagittal plane. Third, signal intensity correction between each data is performed in the localized area, and renal parenchyma is segmented with the ensemble of two 3D segmentation networks based on UNet++ with image and shape patches. The average F1-score of renal parenchyma was 91.43% at left kidney, 89.30% at right kidney in F1-score, respectively.

Paper Type: methodological development

Primary Subject Area: Segmentation

Secondary Subject Area: Application: Other

Paper Status: original work, not submitted yet

Source Code Url: I cannot publish the source code because this research is supported by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Science and ICT.

Data Set Url: I cannot provide the datasets because it belongs to Yonsei University College of Medicine, Seoul, Republic of Korea, not public data.

Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.