Fused Attention Network for Intravascular Optical Coherence Tomography Image Stent Segmentation

Fig. 5. Images of different stent conditions in the dataset. (a) Normal stent image; (b) blood artifact contamination image; (c) stent malapposition image; (d) stent overlap image

Download full size

Fig. 6. Preprocessed images in different coordinate systems; (a) Cartesian coordinate system; (b) polar coordinate system

Download full size

Fig. 7. Visualization of the segmentation results of different algorithms on various types of stent images. (a)‒(c) Normal stent images; (d)‒(f) stent images affected by blood artifacts; (g)‒(i) images of the stent detached from the pipe wall

Download full size

Fig. 8. Visualization of different downsampling effects

Download full size

Fig. 9. Ablation experiment visualization results

Download full size

Table 1. Comparative experimental results of different algorithms

View table

Table 1. Comparative experimental results of different algorithms

Method	IoU	Dice coefficient	Accuracy	Precision	F1-score	Sensitivity
U-Net	61.58	71.91	99.9291	43.89	42.14	44.61
U-Net++	60.95	71.14	99.9317	44.53	41.18	41.15
TransAttUnet	63.46	73.55	99.9315	43.87	42.06	44.29
Attention U-Net	62.47	72.72	99.9311	44.73	42.18	43.93
KiU-Net	63.09	73.15	99.9309	42.79	41.09	42.89
UNeXt	62.67	72.64	99.9307	40.64	40.58	41.99
ConvFormer	64.21	74.05	99.9304	41.83	41.13	45.35
FAU-Net	65.45	75.91	99.9331	44.92	42.28	44.01

Table 2. Comparison results of attention modules
View table
Table 2. Comparison results of attention modules
Module IoU Dice coefficient Accuracy Precision F1-score Sensitivity
U-Net+CBAM 62.36 72.65 99.9311 43.65 42.01 44.08
U-Net+FA 63.46 73.55 99.9315 43.87 42.06 44.29

Table 3. Results of ablation experiments

View table

Table 3. Results of ablation experiments

Model	IoU	Dice coefficient	Accuracy	Precision	F1-score	Sensitivity
U-Net	61.58	71.91	99.9291	43.89	42.14	44.61
U-Net+WTD	63.62	73.39	99.9320	44.53	40.76	43.56
U-Net+FA	63.46	73.55	99.9315	43.87	42.06	44.29
U-Net+MSFM	62.16	72.30	99.9281	44.73	42.18	43.93
U-Net+WTD+FA	64.39	74.53	99.9309	42.79	41.09	45.57
U-Net+FA+MSFM	63.86	73.85	99.9324	40.64	40.58	41.99
U-Net+WTD+MSFM	64.77	74.63	99.9307	44.92	42.28	44.01
U-Net+WTD+FA+MSFM	65.89	76.04	99.9335	44.95	42.48	44.01

Table 4. Ablation experiment results of batch size parameter

View table

Table 4. Ablation experiment results of batch size parameter

Batch size	U-Net	U-Net++	TransAttUnet	Attention U-Net	KiU-Net	UNeXt	FAU-Net
2	71.88	71.15	73.61	72.64	73.11	72.62	75.84
4	71.91	71.14	73.55	72.72	73.15	72.64	75.91
6	71.93	71.08	73.49	72.55	73.09	72.71	75.89
8	71.87	71.19	73.44	72.59	73.24	72.59	75.77

Tools

Get Citation

Copy Citation Text

Shaojiang Wei, Wei Zhang. Fused Attention Network for Intravascular Optical Coherence Tomography Image Stent Segmentation[J]. Laser & Optoelectronics Progress, 2025, 62(16): 1617001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites