Computer Vision News - December 2018

17 Combined (CT-MR) Healthy Abdominal Organ Segmentation transfer learning to fine tune a trained model to abdominal organ segmentation or other strategies such as data augmentation are also encouraged. Competition categories and aim: CHAOS has two aims: segmentation of liver from CT ; and segmentation of four abdominal organs (i.e. liver, spleen, right and left kidneys) from MRI acquired with two sequences (T1- DUAL and T2-SPIR). There will be five competition categories in which the participating teams can take place and submit their result(s): 1. Liver Segmentation (CT- MRI), 2. Liver Segmentation (CT only), 3. Liver Segmentation (MRI only), 4. Segmentation of abdominal organs (CT+MRI), 5. Segmentation of abdominal organs (MRI only). Teams can participate in a single category or can submit for multiple categories using different systems. However, categories 1 and 4 are especially important for us since there already exist very successful models individually for CT and MRI, but not working effectively in both at the same time. FollowUps of CHAOS: CHAOS is not only about finding the best method. Besides providing a comparative study of participating algorithms, the competition will be used to gain insight about complementarity and diversity of different methods. Recent developments show that classifier ensembles can provide higher performance than its components if certain conditions are met. EMMA , which won the BRATS challenge in 2018, is one of the most recent examples. We believe that the outcomes of CHAOS will spark many ideas to improve existing ensemble strategies. Furthermore, it is already being planned to extend the CHAOS to cover vessel tree extraction inside the liver and Coinaud classification to determine the volumes of its sub-segments. Challenge Computer Vision News 2D illustration of the outcomes of different segmentationmethods Red: True Positives, Blue: FalseNegatives, Yellow: False Positives Example annotations for T1-DUAL images