Polyp Segmentation Method Combining HarDNet and Reverse Attention

Ziqi Han; Qiaohong Liu; Chen Ling; Jiawei Liu; Cunjue Liu

doi:10.3788/LOP212665

Journals >Laser & Optoelectronics Progress >Volume 60 >Issue 2 >Page 0217003 > Article

Laser & Optoelectronics Progress
Vol. 60, Issue 2, 0217003 (2023)

Polyp Segmentation Method Combining HarDNet and Reverse Attention

Ziqi Han¹, Qiaohong Liu^2,*, Chen Ling², Jiawei Liu¹, and Cunjue Liu¹

Author Affiliations

¹College of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China

²College of Medical Instruments, Shanghai University of Medicine and Health Sciences, Shanghai 201318, China

show less

DOI: 10.3788/LOP212665 Cite this Article Set citation alerts

Ziqi Han, Qiaohong Liu, Chen Ling, Jiawei Liu, Cunjue Liu. Polyp Segmentation Method Combining HarDNet and Reverse Attention[J]. Laser & Optoelectronics Progress, 2023, 60(2): 0217003 Copy Citation Text

show less

Fig. 1. HraNet model structure

Download full size

Fig. 2. Structure comparison diagram of DenseNet and HarDNet

Download full size

Fig. 3. Reverse attention block

Download full size

Fig. 4. Receptive field block

Download full size

Fig. 5. Comparison of segmentation results of colonic polyps

Download full size

Parameter	m	Stride 2	Stride 4	Stride 8	Stride 16	Stride 32
Value	1.7	3×3，32，stride is 2	8（HDB），k=14，t=128	16（HDB），k=16，t=256	16（HDB），k=40，t=640	4（HDB），k=160，t=1024
Value	1.7	3×3，64	8（HDB），k=14，t=128	16（HDB），k=20，t=320	16（HDB），k=40，t=640	4（HDB），k=160，t=1024

Table 1. Detailed implementation parameters of HarDNet68

Dataset	Method	mDice	mIoU	$F_{β}^{w}$	$S_{α}$	$E_{ϕ}^{m a x}$	MAE
Kvasir SEG	UNet	0.818	0.746	0.794	0.858	0.893	0.055
	UNet++	0.821	0.743	0.808	0.862	0.910	0.048
	SFA	0.723	0.611	0.670	0.782	0.849	0.075
	PraNet	0.898	0.840	0.885	0.915	0.948	0.030
	HraNet	0.901	0.845	0.890	0.917	0.944	0.027
CVC ClinicDB	UNet	0.823	0.755	0.811	0.889	0.954	0.019
	UNet++	0.794	0.729	0.785	0.873	0.931	0.022
	SFA	0.700	0.607	0.647	0.793	0.885	0.042
	PraNet	0.899	0.849	0.896	0.936	0.979	0.009
	HraNet	0.930	0.879	0.928	0.949	0.980	0.008

Table 2. Comparison of segmentation effects of different methods on Kvasir SEG and CVC ClinicDB datasets

Dataset	Method	mDice	mIoU	$F_{β}^{w}$	$S_{α}$	$E_{ϕ}^{m a x}$	MAE
CVC ColonDB	UNet	0.512	0.444	0.498	0.712	0.776	0.061
	UNet++	0.483	0.410	0.467	0.691	0.760	0.064
	SFA	0.469	0.347	0.379	0.634	0.765	0.094
	PraNet	0.709	0.640	0.696	0.819	0.869	0.045
	HraNet	0.729	0.660	0.720	0.830	0.858	0.038
ETISLarib Polyp DB	UNet	0.398	0.335	0.366	0.684	0.740	0.036
	UNet++	0.401	0.344	0.390	0.683	0.776	0.035
	SFA	0.297	0.217	0.231	0.557	0.633	0.109
	PraNet	0.628	0.567	0.600	0.794	0.841	0.031
	HraNet	0.650	0.581	0.617	0.797	0.827	0.035
Endosece	UNet	0.710	0.627	0.684	0.843	0.876	0.022
	UNet++	0.707	0.624	0.687	0.839	0.898	0.018
	SFA	0.467	0.329	0.341	0.640	0.817	0.065
	PraNet	0.871	0.797	0.843	0.925	0.972	0.010
	HraNet	0.892	0.824	0.875	0.937	0.964	0.007

Table 3. Generalization capability test results

Dataset	Method	Epoch	Training /min	Inference /（frame·s^-1）	mDice	mIoU
CVC ClinicDB	UNet	50	~70	~126	0.823	0.755
	UNet++	50	~80	~53	0.794	0.729
	SFA	200	>600	~70	0.700	0.607
	PraNet	50	~75	~115	0.899	0.849
	HraNet	50	~33	~148	0.930	0.879

Table 4. Comparison of training time and reasoning speed of different methods based on same platform

Ziqi Han, Qiaohong Liu, Chen Ling, Jiawei Liu, Cunjue Liu. Polyp Segmentation Method Combining HarDNet and Reverse Attention[J]. Laser & Optoelectronics Progress, 2023, 60(2): 0217003

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information

微信扫一扫：分享

微信扫一扫：分享