Context-aware Attention U-Net for the segmentation of pores in Lamina Cribrosa using partial points annotation - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Context-aware Attention U-Net for the segmentation of pores in Lamina Cribrosa using partial points annotation

Résumé

Glaucoma is the second leading cause of blindness in the world. Although its physiopathology remains unclear, the lamina cribrosa, a 3D mesh-like structure consisting of pores, that allow the axons passing through to join the brain, has been identified as the primary site of damage. In this work we present an extended version of U-Net for pore segmentation in 2D enface OCT images with partial point annotations, i.e. having only a small portion of pore locations in each image labeled. Our method combines the attention gate and the context information to address the difficulties caused by small object segmentation in low signal to noise ratio images. Experimental results show that 71.8% of the annotated pores are successfully segmented.
Fichier principal
Vignette du fichier
ICMLA2022_pore_segmentation_LC_v3.pdf (3.19 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03800461 , version 1 (06-10-2022)

Identifiants

  • HAL Id : hal-03800461 , version 1

Citer

Nan Ding, Hélène Urien, Florence Rossant, Jérémie Sublime, Michel Paques. Context-aware Attention U-Net for the segmentation of pores in Lamina Cribrosa using partial points annotation. 21st IEEE International Conference on Machine Learning and Applications (IEEE ICMLA'22), Dec 2022, Bahamas, Bahamas. ⟨hal-03800461⟩
42 Consultations
42 Téléchargements

Partager

Gmail Facebook X LinkedIn More