Accuracy improvement for classifying retinal OCT images by diseases using deep learning-based selective denoising approach
In ophthalmology, retinal optical coherence tomography (OCT) images with noticeable structural features help identify human eyes as healthy or diseased. The recently hot artificial intelligence (AI) realized this recognition process automatically. However, speckle noise in the original retinal OCT image reduces the accuracy of disease classification. This study presents a time-saving approach based on deep learning to improve classification accuracy by removing the noise from the original dataset. Firstly, four pre-trained convolutional neural networks (CNNs) from the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) were trained to classify the original images into two categories: The noise reduction required (NRR) and the noise-free (NF) images. Among the CNNs, VGG19_BN performed best with 98% accuracy and 99% recall. Then, we used the block-matching and 3D filtering (BM3D) algorithm to denoise the NRR images. Those noise-removed NRR and the NF images form the processed dataset. The quality of images in the dataset is prominently ameliorated after denoising, which is valid to improve the models’ performance. The original and processed datasets were tested on the four pre-trained CNNs to evaluate the effectiveness of our proposed approach. We have compared the CNNs, and the results show the performance of the CNNs trained with the processed dataset is improved by an average of 2.04%, 5.19%, and 5.10% under overall accuracy (OA), Macro F1-score, and Micro F1-score, respectively. Especially for DenseNet161, the OA is improved to 98.14%. Our proposed method demonstrates its effectiveness in improving classification accuracy and opens a new solution to reduce denoising time-consuming for large datasets.
[1] HuangD., SwansonE. A., LinC. P., SchumanJ. S., StinsonW. G., ChangW., HeeM. R., FlotteT., GregoryK., PuliafitoC. A., “Optical coherence tomography,” Science254, 1178–1181 (1991).
[2] ChenT. C., HoguetA., JunkA. K., Nouri-MahdaviK., RadhakrishnanS., TakusagawaH. L., ChenP. P., “Spectral-domain OCT: Helping the clinician diagnose glaucoma: A report by the American Academy of Ophthalmology,” Ophthalmology125, 1817–1827 (2018).
[3] PujariA., BhaskaranK., SharmaP., SinghP., PhuljheleS., SaxenaR., AzadS. V., “Optical coherence tomography angiography in neuro-ophthalmology: Current clinical role and future perspectives,” Surv. Ophthalmol.66, 471–481 (2021).
[4] LiH., LiuK., YaoL., DengX., ZhangZ., LiP., “ID-OCTA: OCT angiography based on inverse SNR and decorrelation features,” J. Innov. Opt. Health Sci.14, 2130001 (2021).
[5] Schmidt-ErfurthU., SadeghipourA., GerendasB. S., WaldsteinS. M., BogunovićH., “Artificial intelligence in retina,” Prog. Retin. Eye Res.67, 1–29 (2018).
[6] KapoorR., WhighamB. T., Al-AswadL. A., “Artificial intelligence and optical coherence tomography imaging,” Asia-Pacific J. Ophthalmol.8, 187–194 (2019).
[7] TingD. S. W., PasqualeL. R., PengL., CampbellJ. P., LeeA. Y., RamanR., TanG. S. W., SchmettererL., KeaneP. A., WongT. Y., “Artificial intelligence and deep learning in ophthalmology,” Br. J. Ophthalmol.103, 167–175 (2019).
[8] OngsuleeP.,Artificial intelligence, machine learning and deep learning, 2017 15th Int. Conf. ICT and Knowledge Engineering (ICT&KE), pp. 1–6, IEEE (2017).
[9] LiuJ., YanS., LuN., YangD., FanC., LvH., WangS., ZhuX., ZhaoY., WangY., “Automatic segmentation of foveal avascular zone based on adaptive watershed algorithm in retinal optical coherence tomography angiography images,” J. Innov. Opt. Health Sci.15, 2242001 (2022).
[10] ZhengG., JiangY., ShiC., MiaoH., YuX., WangY., ChenS., LinZ., WangW., LuF., “Deep learning algorithms to segment and quantify the choroidal thickness and vasculature in swept-source optical coherence tomography images,” J. Innov. Opt. Health Sci.14, 2140002 (2021).
[11] PrahsP., RadeckV., MayerC., CvetkovY., CvetkovaN., HelbigH., MärkerD., “OCT-based deep learning algorithm for the evaluation of treatment indication with anti-vascular endothelial growth factor medications,” Graefe’s Arch. Clin. Exp. Ophthalmol.256, 91–98 (2018).
[12] KermanyD. S., GoldbaumM., CaiW., ValentimC. C., LiangH., BaxterS. L., McKeownA., YangG., WuX., YanF., “Identifying medical diagnoses and treatable diseases by image-based deep learning,” Cell172, 1122–1131. e1129 (2018).
[13] GulshanV., PengL., CoramM., StumpeM. C., WuD., NarayanaswamyA., VenugopalanS., WidnerK., MadamsT., CuadrosJ., “Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs,” Jama316, 2402–2410 (2016).
[14] JiangZ., WangL., WuQ., ShaoY., ShenM., JiangW., DaiC., “Computer-aided diagnosis of retinopathy based on vision transformer,” J. Innov. Opt. Health Sci.15, 2250009 (2022).
[15] KimJ., TranL.,Retinal disease classification from OCT images using deep learning algorithms, 2021 IEEE Conf. Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), pp. 1–6, IEEE (2021).
[16] ChenZ., ZengZ., ShenH., ZhengX., DaiP., OuyangP., “DN-GAN: Denoising generative adversarial networks for speckle noise reduction in optical coherence tomography images,” Biomed. Signal Process. Control55, 101632 (2020).
[17] KoziarskiM., CyganekB., “Image recognition with deep neural networks in presence of noise–dealing with and taking advantage of distortions,” Integr. Comput.-Aided Eng.24, 337–349 (2017).
[18] DabovK., FoiA., KatkovnikV., EgiazarianK., “Image denoising by sparse 3-D transform-domain collaborative filtering,” IEEE Trans. Image Process.16, 2080–2095 (2007).
[19] QiuB., HuangZ., LiuX., MengX., YouY., LiuG., YangK., MaierA., RenQ., LuY., “Noise reduction in optical coherence tomography images using a deep neural network with perceptually-sensitive loss function,” Biomed. Opt. Exp.11, 817–830 (2020).
[20] GholamiP., RoyP., ParthasarathyM. K., LakshminarayananV., “OCTID: Optical coherence tomography image database,” Comput. Electr. Eng.81, 106532 (2020).
[21] LiM., IdoughiR., ChoudhuryB., HeidrichW., “Statistical model for OCT image denoising,” Biomed. Opt. Exp.8, 3903–3917 (2017).
[22] BaumannB., MerkleC. W., LeitgebR. A., AugustinM., WartakA., PircherM., HitzenbergerC. K., “Signal averaging improves signal-to-noise in OCT images: But which approach works best, and when?,” Biomed. Opt. Exp.10, 5755–5775 (2019).
[23] ZhangA., XiJ., SunJ., LiX., “Pixel-based speckle adjustment for noise reduction in Fourier-domain OCT images,” Biomed. Opt. Exp.8, 1721–1730 (2017).
[24] WangJ., DengG., LiW., ChenY., GaoF., LiuH., HeY., ShiG., “Deep learning for quality assessment of retinal OCT images,” Biomed. Opt. Exp.10, 6057–6072 (2019).
[25] LiuX., TanakaM., OkutomiM., “Single-image noise level estimation for blind denoising,” IEEE Trans. Image Process.22, 5226–5237 (2013).
[26] WeissK., KhoshgoftaarT. M., WangD., “A survey of transfer learning,” J. Big Data3, 1–40 (2016).
[27] TorreyL., ShavlikJ., Handbook of Research on Machine Learning Applications and Trends: Algorithms, Methods, and Techniques,Information Science Reference, Hershey, PA (2010).
[28] RussakovskyO., DengJ., SuH., KrauseJ., SatheeshS., MaS., HuangZ., KarpathyA., KhoslaA., BernsteinM., “Imagenet large scale visual recognition challenge,” Int. J. Comput. Vis.115, 211–252 (2015).
[29] MajiD., Santara A., Mitra P., Sheet D., “Ensemble of deep convolutional neural networks for learning to detect retinal vessels in fundus images,” arXiv: 1603.04833 (2016).
[30] HuangG., LiuZ., Van Der MaatenL., WeinbergerK. Q.,Densely connected convolutional networks, Proc. IEEE Conf. Computer Vision and Pattern Recognition, pp. 4700–4708 (IEEE, Honolulu, HI, USA, 2017).
[31] SzegedyC., IoffeS., VanhouckeV., AlemiA. A.,Inception-v4, inception-resnet and the impact of residual connections on learning, Thirty-first AAAI Conf. Artificial Intelligence (AAAI Press, San Francisco, California, USA, 2017), pp. 4278–4284.
[32] HeK., ZhangX., RenS., SunJ.,Deep residual learning for image recognition, Proc. IEEE Conf. Computer Vision and Pattern Recognition, pp. 770–778 (IEEE, Las Vegas, NV, USA, 2016).
[33] HuJ., ShenL., SunG.,Squeeze-and-excitation networks, Pro. IEEE Conf. Computer Vision and Pattern Recognition, pp. 7132–7141 (IEEE, Salt Lake City, UT, USA, 2018).
[34] SimonyanK., Zisserman A., “Very deep convolutional networks for large-scale image recognition,” preprint, arXiv: 1409.1556 (2014).
[35] ChinchorN.,MUC-4 evaluation metrics, Proc. Fourth Message Understanding Conf., pp. 22–29. Morgan Kaufmann (Association for Computational Linguistics, McLean, Virginia, USA, 1992).
[36] LewisD. D., SchapireR. E., CallanJ. P., PapkaR.,Training algorithms for linear text classifiers, Proc. 19th Annual Int. ACM SIGIR Conf. Research and Development in Information Retrieval, pp. 298–306 (Association for Computing Machinery, New York, NY, United States, 1996).
[37] SelvarajuR. R., CogswellM., DasA., VedantamR., ParikhD., BatraD.,Grad-cam: Visual explanations from deep networks via gradient-based localization, Proc. IEEE Int. Conf. Computer Vision, pp. 618–626 (IEEE, Venice, Italy, 2017).
[38] TianC., FeiL., ZhengW., XuY., ZuoW., LinC.-W., “Deep learning on image denoising: An overview,” Neural Netw.131, 251–275 (2020).
Lantian Hu, Ruixiang Guo, Sifan Li, Jing Cao, Qian Liu. Accuracy improvement for classifying retinal OCT images by diseases using deep learning-based selective denoising approach[J]. Journal of Innovative Optical Health Sciences, 2023, 16(6): 2350008.
PDF全文
