Deep learning based registration using spatial gradients and noisy segmentation labels

Théo Estienne (1, 2, 3) , Maria Vakalopoulou (1, 3) , Enzo Battistella (1, 2, 3) , Alexandre Carré (2) , Théophraste Henry (2) , Marvin Lerousseau (1, 3) , Charlotte Robert (2, 4) , Nikos Paragios (5) , Eric Deutsch (2, 4)

1 MICS - Mathématiques et Informatique pour la Complexité et les Systèmes
2 RaMo-IT - Radiothérapie Moléculaire et Innovation Thérapeutique
3 OPIS - OPtimisation Imagerie et Santé
4 IGR - Institut Gustave Roussy
5 TheraPanacea [Paris]

Image registration is one of the most challenging problems in medical image analysis. In the recent years, deep learning based approaches became quite popular, providing fast and performing registration strategies. In this short paper, we summarise our work presented on Learn2Reg challenge 2020. The main contributions of our work rely on (i) a symmetric formulation, predicting the transformations from source to target and from target to source simultaneously, enforcing the trained representations to be similar and (ii) integration of variety of publicly available datasets used both for pretraining and for augmenting segmentation labels. Our method reports a mean dice of 0.64 for task 3 and 0.85 for task 4 on the test sets, taking third place on the challenge. Our code and models are publicly available at https://github.com/TheoEst/abdominal_registration and https://github.com/TheoEst/hippocampus_registration.

Mots clés: deformable registration; deep learning; abdominal CT; hippocampus MRI; noisy segmentations

Domaines: Intelligence artificielle [cs.AI]; Vision par ordinateur et reconnaissance de formes [cs.CV]; Imagerie médicale

Théo Estienne, Maria Vakalopoulou, Enzo Battistella, Alexandre Carré, Théophraste Henry, et al.. Deep learning based registration using spatial gradients and noisy segmentation labels. 2020. ⟨hal-02974930⟩