Double-path parallel convolutional neural network for removing speckle noise in different types of OCT images

Zhengjie Shen; Manhui Xi; Chen Tang; Min Xu; Zhenkun Lei

doi:10.1364/AO.419871

Applied Optics
Vol. 60,
Issue 15,
pp. 4345-4355
(2021)
•https://doi.org/10.1364/AO.419871

Double-path parallel convolutional neural network for removing speckle noise in different types of OCT images

Zhengjie Shen, Manhui Xi, Chen Tang, Min Xu, and Zhenkun Lei

Not Accessible

Your library or personal account may give you access

Get PDF
Email
Share
Get Citation
Copy Citation Text
Zhengjie Shen, Manhui Xi, Chen Tang, Min Xu, and Zhenkun Lei, "Double-path parallel convolutional neural network for removing speckle noise in different types of OCT images," Appl. Opt. 60, 4345-4355 (2021)

Export Citation
- BibTex
- Endnote (RIS)
- HTML
- Plain Text
Citation alert
Save article

Check for updates

Abstract

Speckle noises widely exist in optical coherence tomography (OCT) images. We propose an improved double-path parallel convolutional neural network (called DPNet) to reduce speckles. We increase the network width to replace the network depth to extract deeper information from the original OCT images. In addition, we use dilated convolution and residual learning to increase the learning ability of our DPNet. We use 100 pairs of human retinal OCT images as the training dataset. Then we test the DPNet model for denoising speckles on four different types of OCT images, mainly including human retinal OCT images, skin OCT images, colon crypt OCT images, and quail embryo OCT images. We compare the DPNet model with the adaptive complex diffusion method, the curvelet shrinkage method, the shearlet-based total variation method, and the OCTNet method. We qualitatively and quantitatively evaluate these methods in terms of image smoothness, structural information protection, and edge clarity. Our experimental results prove the performance of the DPNet model, and it allows us to batch and quickly process different types of poor-quality OCT images without any parameter fine-tuning under a time-constrained situation.

Full Article | PDF Article

More Like This

Reducing speckle in anterior segment optical coherence tomography images based on a convolutional neural network

Liansheng Liu, Zuopan Zhai, Ting Zhang, and Licheng Fan
Appl. Opt. 60(35) 10964-10974 (2021)

Speckle noise reduction in optical coherence tomography images based on edge-sensitive cGAN

Yuhui Ma, Xinjian Chen, Weifang Zhu, Xuena Cheng, Dehui Xiang, and Fei Shi
Biomed. Opt. Express 9(11) 5129-5146 (2018)

Unsupervised despeckling of optical coherence tomography images by combining cross-scale CNN with an intra-patch and inter-patch based transformer

Quan Zhou, Mingwei Wen, Mingyue Ding, and Xuming Zhang
Opt. Express 30(11) 18800-18820 (2022)

Previous Article Next Article

Data Availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Figures (13)

You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Tables (8)

You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Equations (15)

You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

	Curvelet	ACD						TV-SH
OCT	Scales	$T$	$K_{min}$	$K_{max}$	$θ$	$a$	$b$	$μ$	$λ$	$η^{'}$	$η^{''}$
OCT-1	5	3.0	2	28	$π / 30$	0.20	0.70	1	1	0.10	100
OCT-2	6	4.0	2	28	$π / 30$	0.25	0.75	1	1	0.12	100
OCT-3	6	4.0	2	28	$π / 30$	0.20	0.70	1	1	0.11	100
OCT-4	5	3.0	2	28	$π / 30$	0.30	0.75	1	1	0.13	100
OCT-5	5	5.0	2	28	$π / 30$	0.30	0.75	1	1	0.15	100
OCT-6	5	4.0	2	28	$π / 30$	0.20	0.60	1	1	0.15	100
OCT-7	3	3.0	2	28	$π / 30$	0.25	0.75	1	1	0.10	100
OCT-8	4	4.0	2	28	$π / 30$	0.25	0.70	1	1	0.10	100

	STV				SOOPDE		FCM
OCT	$μ$	$λ$	$η^{'}$	$η^{''}$	$Δ t$	$N$	$p_{1}$	$p_{2}$	$p_{3}$	$p_{4}$
OCT-1	1	1	0.10	100	0.12	70	50	500	0.001	1
OCT-2	1	1	0.12	100	0.20	100	60	500	0.001	1
OCT-3	1	1	0.11	100	0.15	50	45	500	0.001	1
OCT-4	1	1	0.13	100	0.14	150	70	500	0.001	1

Methods		SNR	CNR	ENL	PSNR	SSIM	Time
Original	Mean	28.60	4.37	258.49	22.77	0.22	–
Original	Std	1.34	0.80	131.41	0.48	0.035	–
Curvelet	Mean	39.95	9.46	495.26	23.87	0.35	5.2
Curvelet	Std	2.12	1.05	214.58	0.45	0.022	–
ACD	Mean	43.86	6.83	394.09	23.27	0.42	7.7
ACD	Std	1.22	1.05	240.96	0.47	0.027	–
TV-SH	Mean	37.31	11.40	984.71	27.15	0.71	14.7
TV-SH	Std	2.73	2.33	273.53	0.38	0.035	–
OCTNet	Mean	50.46	16.78	1031.83	28.32	0.81	0.14
OCTNet	Std	2.91	2.39	297.17	0.49	0.038	–
DPNet	Mean	74.99	20.86	1539.82	29.75	0.89	0.15
DPNet	Std	2.31	2.39	293.62	0.67	0.063	–

Methods	PSNR	SSIM
Upper network with one RL	29.65	0.84
Lower network with one RL	29.61	0.83
Double-path network with dilation convolution and one RL	29.70	0.85
Double-path network with all ordinary convolution and three RL	29.03	0.83
Double-path network with dilation convolution and three RL	29.75	0.89

OCT-5	SNR	CNR	ENL	Time
Original	25.2728	1.9573	12.6556	–
Curvelet	32.3633	2.8436	28.7913	3.5
ACD	38.1611	2.0314	14.8690	6.1
TV-SH	28.2013	2.8852	26.6631	19.3
OCTNet	41.3803	4.1374	31.4767	0.13
DPNet	51.4869	3.5433	33.0238	0.14

OCT-6	SNR	CNR	ENL	Time
Original	25.6863	2.6794	36.4545	–
Curvelet	34.1976	4.2446	88.0384	4.1
ACD	38.0025	3.0967	64.4827	4.9
TV-SH	33.3390	4.1648	80.4340	18.1
OCTNet	39.7936	5.4729	107.0409	0.14
DPNet	53.4940	6.8123	180.1658	0.14

OCT-7	SNR	CNR	ENL	Time
Original	20.2021	1.5772	17.2984	–
Curvelet	28.7617	2.7830	49.0954	3.4
ACD	32.1681	2.4115	20.8879	5.8
TV-SH	29.2721	2.8897	56.9151	15.6
OCTNet	33.7719	4.1959	81.5582	0.14
DPNet	42.7681	5.1551	100.7822	0.14

OCT-8	SNR	CNR	ENL	Time
Original	24.0839	2.3600	47.7006	–
Curvelet	29.8104	4.3815	189.3632	4.3
ACD	35.8572	3.5428	61.7997	5.8
TV-SH	27.1808	3.7474	173.6289	19.2
OCTNet	30.1331	5.6703	211.8486	0.14
DPNet	42.9365	9.6154	381.4953	0.14

Abstract

Data Availability

Cited By

Figures (13)

Tables (8)

Equations (15)

Applied Optics