Models for the participation to the Learn2Reg 2020 Challenge (Task 3 & 4)

In this upload, we give our models used for the participation to the Learn2Reg 2020 Challenge. We participated to 2 tasks : abdominal registration (task 3) and hippocampus registration (task 4) and for each task different models are given with different hyperparameters. Our method reports a mean dice of 0.64 for task 3 and 0.85 for task 4 on the test sets, taking 3rd place on the challenge. More details are given in article and github repository. Description of the training parameters is available in the article : Deep learning based registration using spatialgradients and noisy segmentation labels (http://128.84.4.34/pdf/2010.10897) The code to use and run these models is available in the two following GitHub repository : https://github.com/TheoEst/abdominal_registration https://github.com/TheoEst/hippocampus_registration Full affiliation are : Théeo Estienne12, Maria Vakalopoulou1, Enzo Battistella12, Alexandre Carré2, Théophraste Henry2, Marvin Lerousseau12, Charlotte Robert2, Nikos Paragios3 and Eric Deutsch2 1 Université Paris-Saclay, CentraleSupélec, Mathématiques et Informatique pour la Complexité et les Systèmes, 91190, Gif-sur-Yvette, France. 2 Université Paris-Saclay, Institut Gustave Roussy, Inserm, Radiothérapie Moléculaire et Innovation Thérapeutique, 94800, Villejuif, France. 3 Therapanace