Abstract
Annotating retinal layers in optical coherence tomography (OCT) images is a time-consuming and costly process, requiring expert knowledge. Weakly supervised segmentation methods that automatically discover layers in OCT images can alleviate this problem and enable the analysis of large datasets from new devices without the need for costly annotations. We propose a novel approach to weakly supervised retinal layer segmentation by incorporating principles from object-centric learning into the Spatial Decomposition Network framework. Our model learns to segment retinal layers requiring only annotations for the internal limiting membrane (ILM) and the outer boundary of the retinal pigment epithelium (OB_RPE). Fluid segmentation is supported as an optional extension when fluid labels are available for training. We demonstrate its ability to generalize across different devices on a large and diverse set of OCT datasets and compare it against existing (weakly) supervised methods.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/OMIA_049.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=si7zLCO441
BibTex
@InProceedings{SeiMar_Weakly_MICCAISAT2026,
author = { Seibel, Marc S. AND Handels, Heinz AND Ehrhardt, Jan},
title = { { Weakly Supervised Retinal Layer Segmentation in OCT } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17270},
month = {pending},
page = {pending}
}