notesum.ai

Published at December 6

Uncertainty-aware retinal layer segmentation in OCT through probabilistic signed distance functions

eess.IV

cs.AI

cs.CV

Released Date: December 6, 2024

Authors: Mohammad Mohaiminul Islam¹, Coen de Vente¹, Bart Liefers², Caroline Klaver², Erik J Bekkers¹, Clara I. Sánchez¹

Aff.: ¹University of Amsterdam, Netherlands; ²Erasmus MC, Rotterdam, Netherlands

Arxiv: http://arxiv.org/pdf/2412.04935v1

Refer to caption

Method	Layer	MAE
		(internal)	(external)
	ILM	$0.97_{\pm 0.72}$	$2.37_{\pm 1.83}$
REGR	RPE	$1.32_{\pm 0.51}$	$2.54_{\pm 2.14}$
	BM	$1.76_{\pm 1.48}$	$5.28_{\pm 2.94}$
	Avg.	1.35	3.40
	ILM	$0.85_{\pm 0.71}$	$2.98_{\pm 2.41}$
pREGR	RPE	$1.28_{\pm 0.82}$	$2.71_{\pm 2.00}$
	BM	$1.80_{\pm 1.50}$	$3.69_{\pm 1.78}$
	Avg.	1.31	3.12
	ILM	$0.33_{\pm 0.42}$	$1.61_{\pm 0.75}$
SDF (Our)	RPE	$0.60_{\pm 0.41}$	$1.02_{\pm 1.64}$
	BM	$0.79_{\pm 0.42}$	$1.39_{\pm 1.01}$
	Avg.	0.57	1.34
	ILM	$0.32_{\pm 0.44}$	$1.63_{\pm 0.78}$
pSDF (Our)	RPE	$0.59_{\pm 0.40}$	$1.02_{\pm 1.47}$
	BM	$0.75_{\pm 0.29}$	$1.27_{\pm 0.82}$
	Avg.	0.55	1.30