Table 3 Ablation studies of different implementations of AFB on the DDR dataset, where [self, cross] denotes the simple concatenation of the self-attention mechanism and cross-attention mechanism over the channels, concat. denotes the concatenation of the lesion-feature based and vascular-feature based keys over the channels, and + denotes the spatial summation of these two keys.

From: Prior-guided attention fusion transformer for multi-lesion segmentation of diabetic retinopathy

Fusion method

[Self, Cross]

AFB

AUPR

mAUPR

  

\(K_l\)

\(K_v\)

EXs

HEs

MAs

SEs

 

N/A

\(\checkmark\)

  

0.5522

0.2302

0.1235

0.2428

0.2872

N/A

 

\(\checkmark\)

 

0.5473

0.455

0.1725

0.3144

0.3723

Concat.

 

\(\checkmark\)

\(\checkmark\)

0.5499

0.4413

0.1623

0.3316

0.3713

+(ours)

 

\(\checkmark\)

\(\checkmark\)

0.5724

0.4256

0.202

0.3608

0.3902

  1. Best: bold.