notesum.ai

Published at November 22

Boundless Across Domains: A New Paradigm of Adaptive Feature and Cross-Attention for Domain Generalization in Medical Image Segmentation

cs.CV

cs.AI

Released Date: November 22, 2024

Authors: Yuheng Xu¹, Taiping Zhang¹

Aff.: ¹College of Computer Science, Chongqing University

Arxiv: http://arxiv.org/abs/2411.14883v1

Refer to caption

Task	Optic Cup/Disc Segmentation					Optic Cup/Disc Segmentation
Task	Domain 1	Domain 2	Domain 3	Domain 4	Avg	Domain 1	Domain 2	Domain 3	Domain 4	Avg
DoFE	79.66/95.95	79.62/88.42	85.45/90.53	83.68/94.84	87.27	20.85/7.04	14.44/18.20	10.22/13.20	9.11/5.81	12.36
HCDG	78.26/92.69	80.14/90.07	84.88/92.88	86.27/94.72	87.49	21.94/12.31	14.55/14.65	10.42/10.05	7.87/5.85	12.21
TVConv	85.87/96.28	78.17/91.72	86.86/92.19	82.04/91.95	88.14	14.78/6.51	15.60/14.84	9.33/10.95	10.17/10.13	11.54
RAM-DSIR	83.08/95.59	76.98/87.30	84.91/94.86	84.45/93.65	87.60	17.92/7.90	16.99/18.84	10.51/7.38	9.18/7.49	12.03
CDDSA	85.78/96.36	78.38/91.81	86.26/95.10	83.18/92.59	88.68	14.84/6.33	14.48/11.84	9.73/7.04	9.68/8.13	10.26
WT-PSE	86.71/96.23	79.64/92.67	87.90/93.69	83.75/93.04	89.20	14.10/6.76	13.46/10.50	8.97/8.84	9.49/8.85	10.12
Ours	85.86/95.89	80.76/89.42	86.46/94.96	87.62/94.58	89.44	13.29/8.15	13.58/8.93	8.08/8.29	7.19/5.96	9.18