Deepseg: Abdominal Organ Segmentation Using Deep Convolutional Neural Networks
Other conference contribution, 2016

A fully automatic method for abdominal organ segmentation is presented. The method uses a robust initialization step based on a multi-atlas approach where the center of the organ is estimated together with a region of interest surrounding the center. As a second step a convolutional neural network performing pixelwise classification is applied. The convolutional neural network consists of several full 3D convolutional layers and takes two input features, which are designed to ensure both local and global consistency. Despite limited training data, our preliminary experimental results are on par with state-of-the-art approaches that have been developed over many years. More specifically the method is applied to the MICCAI2015 challenge “Multi-Atlas Labeling Beyond the Cranial Vault” in the free competition for organ segmentation in the abdomen. It achieved the best results for 3 out of the 13 organs with a total mean dice coefficient of 0.757 for all organs. Top score was achieved for the gallbladder, the aorta and the right adrenal gland.


Convolutional Neural Networks

Medical Image Analysis


Måns Larsson

Chalmers, Signals and Systems

Yuhang Zhang

Chalmers, Signals and Systems

Fredrik Kahl

Chalmers, Signals and Systems, Signal Processing and Biomedical Engineering


Subject Categories


Medical Image Processing

More information