Semi-supervised Domain Adaptation for Automatic Quality Control of FLAIR MRIs in a Clinical Data Warehouse

Sophie Loizillon (1, 2, 3) , Olivier Colliot (1, 3) , Lydia Chougar (1) , Sebastian Stroer (4) , Yannick Jacob (5) , Aurélien Maire (5) , Didier Dormont (4, 3) , Ninon Burgos (3)

1 ICM - Institut du Cerveau = Paris Brain Institute
2 SU - Sorbonne Université
3 ARAMIS - Algorithms, models and methods for images and signals of the human brain = Algorithmes, modèles et méthodes pour les images et les signaux du cerveau humain [ICM Paris]
4 CHU Pitié-Salpêtrière [AP-HP]
5 AP-HP - Assistance publique - Hôpitaux de Paris (AP-HP)

Domain adaptation is a very useful approach to exploit the potential of clinical data warehouses, which gather a vast amount of medical imaging encompassing various modalities, sequences, manufacturers and machines. In this study, we propose a semi-supervised domain adaptation (SSDA) framework for automatically detecting poor quality FLAIR MRIs within a clinical data warehouse. Leveraging a limited number of labelled FLAIR and a large number of labelled T1-weighted MRIs, we introduce a novel architecture based on the well known Domain Adversarial Neural Network (DANN) that incorporates a specific classifier for the target domain. Our method effectively addresses the covariate shift and class distribution shift between T1-weighted and FLAIR MRIs, surpassing existing SSDA approaches by more than 10 percent points.

Mots clés: Domain Adaptation; Deep Learning; Magnetic Resonance Imaging; Clinical Data Warehouse

Domaines: Informatique [cs]

Sophie Loizillon, Olivier Colliot, Lydia Chougar, Sebastian Stroer, Yannick Jacob, et al.. Semi-supervised Domain Adaptation for Automatic Quality Control of FLAIR MRIs in a Clinical Data Warehouse. DART 2023 - 5th MICCAI Workshop on Domain Adaptation and Representation Transfer, Oct 2023, Vancouver (BC), Canada. pp.84-93, ⟨10.1007/978-3-031-45857-6_9⟩. ⟨hal-04273997⟩