Deep learning for diabetic retinopathy screening, classification and management

doi:10.3877/cma.j.issn.2095-2007.2023.04.010

Abstract

Abstract:

Artificial intelligence technology (AI) based on deep learning (DL) has shown good results and great potential in the application research of diabetes retinopathy (DR). Combining with telemedicine technology can significantly improve the work efficiency for grassroots screening and monitoring DR patients. At present, the application of DL in DR research still faces issues such as matching clinical needs, interpretability, technical bottlenecks, and the acceptance of DL systems by physicians and patients. The research progress of DL technology in DR screening, rating, prediction, and optimization management in recent years was reviewed in this paper.

Key words: Diabetic retinopathy, Artificial intelligence, Deep learning, Screening, Classification, Prediction

Jingke Li, Yanchun Zhang, Jiayi Wu, Xiuyu Ren. Deep learning for diabetic retinopathy screening, classification and management[J]. Chinese Journal of Ophthalmologic Medicine(Electronic Edition), 2023, 13(04): 241-246.

/ / Recommend

Add to citation manager EndNote|Ris|BibTeX

URL: https://zhykyxzz.cma-cmc.com.cn/EN/10.3877/cma.j.issn.2095-2007.2023.04.010

https://zhykyxzz.cma-cmc.com.cn/EN/Y2023/V13/I04/241

Figures/Tables 4

References 49

[1]	Sun H, Saeedi P, Karuranga S, et al. IDF Diabetes Atlas: Global, regional and country－level diabetes prevalence estimates for 2021 and projections for 2045[J]. Diabetes Res Clin Pract, Jan 2022, 183: 109－119.
[2]	Teo ZL, Tham YC, Yu M, et al. Do we have enough ophthalmologists to manage vision－threatening diabetic retinopathy？[J]. Eye, 2020, 34(7): 1255－1261.
[3]	Tan G, Cheung N, Simo R, et al. Diabetic macular oedema[J]. Lancet Diabetes Endocrinol, 2017, 5(2): 143－155.
[4]	Wilkinson CP, Ferris FL, Klein RE, et al. Proposed international clinical diabetic retinopathy and diabetic macular edema disease severity scales[J]. Ophthalmology, 2003, 110(9): 1677－1682.
[5]	Bresnick GH, Mukamel DB, Dickinson JC, et al. A screening approach to the surveillance of patients with diabetes for the presence of vision－threatening retinopathy[J]. Ophthalmology, 2000, 107(1): 19－24.
[6]	Wong TY, Sabanayagam C. Strategies to tackle the global burden of diabetic retinopathy: from epidemiology to artificial intelligence[J]. Ophthalmologica, 2020, 243(1): 9－20.
[7]	Vujosevic S, Aldington SJ, Silva P, et al. Screening for diabetic retinopathy: new perspectives and challenges[J]. The Lancet Diabetes & Endocrinology, 2020, 8(4): 337－347.
[8]	Gunasekeran DV, Ting DS, Tan GS, et al. Artificial intelligence for diabetic retinopathy screening, prediction and management[J]. Current opinion in ophthalmology, 2020, 31(5): 357－365.
[9]	范雯，王晓玲，马枭，等. 基于超广角荧光素眼底血管造影图像行糖尿病视网膜病变分期的多模态深度学习模型研究[J]. 中华眼底病杂志，2022，38(2)：139－145.
[10]	段恺睿，张弘. 人工智能在眼科光相干断层扫描图像中的应用[J]. 中华实验眼科杂志，2022，40(1)：83－87.
[11]	Wang S, Yin Y, Cao G, et al. Hierarchical retinal blood vessel segmentation based on feature and ensemble learning[J]. Neurocomputing, 2015, 149: 708－717.
[12]	Kauppi T, Kalesnykiene V, Kämäräinen JK, et al. DIARETDB 0: Evaluation Database and Methodology for Diabetic Retinopathy Algorithms[G]Paper presented at: Proceedings of the British Machine Vision Conference 2007, University of Warwick, UK, September, 2007: 1－9.
[13]	Kauppi T, Kalesnykiene V, Kamarainen JK, et al. DIARETDB1 diabetic retinopathy database and evaluation protocol [G]. Paper presented at: Proceedings of the British Machine Vision Conference 2007, University of Warwick, UK, September, 2007: 10－13.
[14]	Prentašiĉ P, Lonĉariĉ S, Vatavuk Z, et al. Diabetic Retinopathy Image Database(DRiDB): A new database for diabetic retinopathy screening programs research[J]. 2013: 704－709.
[15]	Decenciere E, Cazuguel G, Zhang X, et al. TeleOphta: Machine learning and image processing methods for teleophthalmology[J]. Irbm, 2013, 34(2): 196－203.
[16]	Li X, Pang T, Xiong B, et al. Convolutional neural networks based transfer learning for diabetic retinopathy fundus image classification[G]. Paper presented at: 2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP－BMEI), 2017.
[17]	Naqvi SAG, Zafar MF, Haq I. Referral system for hard exudates in eye fundus[J]. Computers in biology and medicine, 2015, 64: 217－235.
[18]	Decencière E, Zhang X, Cazuguel G, et al. Feedback on a publicly distributed image database: the Messidor database[J]. Image Analysis & Stereology, 2014, 33(3): 231－234.
[19]	Porwal P, Pachade S, Kamble R, et al. Indian diabetic retinopathy image dataset (IDRiD): a database for diabetic retinopathy screening research[J]. Data, 2018, 3(3): 25.
[20]	Li T, Gao Y, Wang K, et al. Diagnostic assessment of deep learning algorithms for diabetic retinopathy screening[J]. Information Sciences, 2019, 501: 511－522.
[21]	Mo J, Zhang L. Multi－level deep supervised networks for retinal vessel segmentation[J]. International journal of computer assisted radiology and surgery, 2017, 12(12): 2181－2193.
[22]	Fraz MM, Remagnino P, Hoppe A, et al. An ensemble classification－based approach applied to retinal blood vessel segmentation[J]. IEEE Transactions on Biomedical Engineering, 2012, 59(9): 2538－2548.
[23]	Niemeijer M, Van Ginneken B, Cree MJ, et al. Retinopathy online challenge: automatic detection of microaneurysms in digital color fundus photographs[J]. IEEE transactions on medical imaging, 2009, 29(1): 185－195.
[24]	Arunkumar R, Karthigaikumar P. Multi－retinal disease classification by reduced deep learning features[J]. Neural Computing and Applications, 2017, 28(2): 329－334.
[25]	Hajeb Mohammad Alipour S, Rabbani H, Akhlaghi MR. Diabetic retinopathy grading by digital curvelet transform[J]. Comput Math Methods Med, 2012: 761901.
[26]	Hajeb Mohammad Alipour S, Rabbani H, Akhlaghi M. A new combined method based on curvelet transform and morphological operators for automatic detection of foveal avascular zone[J]. Signal, Image and Video Processing, 2014, 8(2): 205－222.
[27]	Gholami P, Roy P, Parthasarathy MK, et al. OCTID: Optical coherence tomography image database[J]. Computers & Electrical Engineering, 2020, 81: 106532.
[28]	Díaz M, Novo J, Cutrín P, et al. Automatic segmentation of the foveal avascular zone in ophthalmological OCT－A images[J]. PLoS One, 2019, 14(2): e0212364.
[29]	Adem K. Exudate detection for diabetic retinopathy with circular Hough transformation and convolutional neural networks[J]. Expert Systems with Applications, 2018, 114: 289－295.
[30]	Lam C, Yi D, Guo M, et al. Automated detection of diabetic retinopathy using deep learning[J]. AMIA summits on translational science proceedings, 2018: 147.
[31]	Abràmoff MD, Lou Y, Erginay A, et al. Improved automated detection of diabetic retinopathy on a publicly available dataset through integration of deep learning[J]. Investigative ophthalmology & visual science, 2016, 57(13): 5200－5206.
[32]	García G, Gallardo J, Mauricio A, et al. Detection of diabetic retinopathy based on a convolutional neural network using retinal fundus images[G], 2017.
[33]	Zhang W, Zhong J, Yang S, et al. Automated identification and grading system of diabetic retinopathy using deep neural networks[J]. Knowledge－Based Systems, 2019, 175: 12－25.
[34]	Hamwood J, Alonso－Caneiro D, Read SA, et al. Effect of patch size and network architecture on a convolutional neural network approach for automatic segmentation of OCT retinal layers[J]. Biomedical optics express, 2018, 9(7): 3049－3066.
[35]	Yang D, Sun Z, Shi J, et al. A multitask deep－learning system for assessment of diabetic macular ischemia on optical coherence tomography angiography images[J]. Retina, 2022, 42(1): 184－194.
[36]	Abràmoff MD, Lavin PT, Birch M, et al. Pivotal trial of an autonomous AI－based diagnostic system for detection of diabetic retinopathy in primary care offices[J]. NPJ digital medicine, 2018, 1(1): 1－8.
[37]	Rajalakshmi R, Subashini R, Anjana RM, et al. Automated diabetic retinopathy detection in smartphone－based fundus photography using artificial intelligence[J]. Eye, 2018, 32(6): 1138－1144.
[38]	Ribeiro L, Oliveira CM, Neves C, et al. Screening for Diabetic Retinopathy in the Central Region of Portugal. Added Value of Automated 'Disease/No Disease'Grading[J]. Ophthalmologica, 2015, 233(2): 96－103.
[39]	Xie Y, Nguyen QD, Hamzah H, et al. Artificial intelligence for teleophthalmology－based diabetic retinopathy screening in a national programme: an economic analysis modelling study[J]. The Lancet Digital Health, 2020, 2(5): e240－e249.
[40]	Ting DSW, Cheung CYL, Lim G, et al. Development and validation of a deep learning system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with diabetes[J]. Jama, 2017, 318(22): 2211－2223.
[41]	Gulshan V, Rajan RP, Widner K, et al. Performance of a deep－learning algorithm vs manual grading for detecting diabetic retinopathy in India[J]. JAMA ophthalmology, 2019, 137(9): 987－993.
[42]	Ruamviboonsuk P, Krause J, Chotcomwongse P, et al. Deep learning versus human graders for classifying diabetic retinopathy severity in a nationwide screening program[J]. NPJ digital medicine, 2019, 2(1): 1－9.
[43]	Li Z, Keel S, Liu C, et al. An automated grading system for detection of vision－threatening referable diabetic retinopathy on the basis of color fundus photographs[J]. Diabetes care, 2018, 41(12): 2509－2516.
[44]	Bellemo V, Lim G, Rim TH, et al. Artificial intelligence screening for diabetic retinopathy: the real－world emerging application[J]. Current Diabetes Reports, 2019, 19(9): 1－12.
[45]	Hansen MB, Tang H, Wang S, et al. Automated detection of diabetic retinopathy in three European populations[J]. Journal of Clinical & Experimental Ophthalmology, 2016, 7(4): 170－178.
[46]	Son J, Shin JY, Kim HD, et al. Development and validation of deep learning models for screening multiple abnormal findings in retinal fundus images[J]. Ophthalmology, 2020, 127(1): 85－94.
[47]	Hwang DK, Yu WK, Lin TC, et al. Smartphone－based diabetic macula edema screening with an offline artificial intelligence[J]. Journal of the Chinese Medical Association, 2020, 83(12): 1102－1106.
[48]	Rasti R, Allingham MJ, Mettu PS, et al. Deep learning－based single－shot prediction of differential effects of anti－VEGF treatment in patients with diabetic macular edema[J]. Biomedical Optics Express, 2020, 11(2): 1139－1152.
[49]	Zhang L, Yuan M, An Z, et al. Prediction of hypertension, hyperglycemia and dyslipidemia from retinal fundus photographs via deep learning: A cross－sectional study of chronic diseases in central China[J]. PloS one, 2020, 15(5): e0233166.

严重程度评级	散瞳后眼底镜下所见
DR严重程度量表
0＝无明显DR	无异常表现
1＝轻度NPDR	仅见微血管瘤
2＝中度NPDR	除微血管瘤外病灶，但病情较重度NPDR轻
3＝重度NPDR	病灶符合"4：2：1"法则中任一条：四个象限任一象限视网膜出血超过20处，2个及以上象限有静脉串珠改变，1个及以上象限有视网膜内微血管异常但尚无PDR表现
4＝ PDR	出现以下一条或两条：新生血管区，玻璃体/视网膜前出血
DME严重程度量表
无明显DME	后极部无明显的视网膜增厚及硬性渗出
有明显的DME	后极部有明显的视网膜增厚及硬性渗出
轻度DME	视网膜后极增厚或硬渗出，但远离黄斑中心
中度DME	视网膜增厚或硬渗出物接近黄斑中心，但不累及该中心
重度DME	视网膜增厚或硬渗出物累及黄斑的中心

严重程度评级	散瞳后眼底镜下所见
DR严重程度量表
0＝无明显DR	无异常表现
1＝轻度NPDR	仅见微血管瘤
2＝中度NPDR	除微血管瘤外病灶，但病情较重度NPDR轻
3＝重度NPDR	病灶符合"4：2：1"法则中任一条：四个象限任一象限视网膜出血超过20处，2个及以上象限有静脉串珠改变，1个及以上象限有视网膜内微血管异常但尚无PDR表现
4＝ PDR	出现以下一条或两条：新生血管区，玻璃体/视网膜前出血
DME严重程度量表
无明显DME	后极部无明显的视网膜增厚及硬性渗出
有明显的DME	后极部有明显的视网膜增厚及硬性渗出
轻度DME	视网膜后极增厚或硬渗出，但远离黄斑中心
中度DME	视网膜增厚或硬渗出物接近黄斑中心，但不累及该中心
重度DME	视网膜增厚或硬渗出物累及黄斑的中心

技术方法	适宜检测DR病变	优点及适用阶段	缺点
直接检眼镜；裂隙灯显微镜+前置镜；间接检眼镜	HE，EX，CWS，纤维血管膜，TRD，视网膜裂孔，视盘异常，明显的MA、NVE、NVD、DME、IRMA等血管异常	适合移动，价格低；适用于DR全程检查	需散瞳；对小的IRMA敏感度低；无法进行回顾性的审核；需要专业眼科医师，难以满足DR筛查需求及保证准确性
FP	HE，EX，CWS，纤维血管膜，TRD，视网膜裂孔，视盘异常，明显的MA、NVE、NVD、DME、IRMA等血管异常	可随时对眼底表现进行评估；大量应用于DR筛查及监测	可能需要散瞳；检查范围有限
SLO	HE，EX，CWS，纤维血管膜，TRD，视网膜裂孔，视盘异常，明显的MA、NVE、NVD、DME、IRMA等血管异常	检查评估范围大，利于发现周边病变	昂贵；无法在基层推广
FFA	视网膜血管循环，血管渗漏，NPA，IRMA，NVE，NVD，DME，FAZ	用于评估可能危及视力DR的严重程度、全视网膜光凝治疗效果及黄斑水肿的激光治疗	有创性检查，需要对全身状况进行评估；检查范围有限；无法在基层推广
UWFA	视网膜血管循环，血管渗漏，NPA，IRMA，NVE，NVD，DME，FAZ	同FFA，但检查评估范围更大	有创性检查；昂贵；无法在基层推广
OCTA	FAZ、MA、IRMA、NVE、NVD	无创；分层评估视网膜及脉络膜血管形态及灌注情况	无血流信号的血管无法显影；无法判断血管是否渗漏；能显示的视网膜范围尚有限；昂贵，难以在基层推广普及
OCT	视网膜增厚，视网膜内水肿或囊腔、高反射点，视网膜下液，黄斑前膜，玻璃体黄斑牵拉	无创；评估及动态监测DME的最佳方法；评估玻璃体视网膜界面的最佳方法；有时用于DR早期筛查	可能需要散瞳；昂贵，难以在基层推广普及

技术方法	适宜检测DR病变	优点及适用阶段	缺点
直接检眼镜；裂隙灯显微镜+前置镜；间接检眼镜	HE，EX，CWS，纤维血管膜，TRD，视网膜裂孔，视盘异常，明显的MA、NVE、NVD、DME、IRMA等血管异常	适合移动，价格低；适用于DR全程检查	需散瞳；对小的IRMA敏感度低；无法进行回顾性的审核；需要专业眼科医师，难以满足DR筛查需求及保证准确性
FP	HE，EX，CWS，纤维血管膜，TRD，视网膜裂孔，视盘异常，明显的MA、NVE、NVD、DME、IRMA等血管异常	可随时对眼底表现进行评估；大量应用于DR筛查及监测	可能需要散瞳；检查范围有限
SLO	HE，EX，CWS，纤维血管膜，TRD，视网膜裂孔，视盘异常，明显的MA、NVE、NVD、DME、IRMA等血管异常	检查评估范围大，利于发现周边病变	昂贵；无法在基层推广
FFA	视网膜血管循环，血管渗漏，NPA，IRMA，NVE，NVD，DME，FAZ	用于评估可能危及视力DR的严重程度、全视网膜光凝治疗效果及黄斑水肿的激光治疗	有创性检查，需要对全身状况进行评估；检查范围有限；无法在基层推广
UWFA	视网膜血管循环，血管渗漏，NPA，IRMA，NVE，NVD，DME，FAZ	同FFA，但检查评估范围更大	有创性检查；昂贵；无法在基层推广
OCTA	FAZ、MA、IRMA、NVE、NVD	无创；分层评估视网膜及脉络膜血管形态及灌注情况	无血流信号的血管无法显影；无法判断血管是否渗漏；能显示的视网膜范围尚有限；昂贵，难以在基层推广普及
OCT	视网膜增厚，视网膜内水肿或囊腔、高反射点，视网膜下液，黄斑前膜，玻璃体黄斑牵拉	无创；评估及动态监测DME的最佳方法；评估玻璃体视网膜界面的最佳方法；有时用于DR早期筛查	可能需要散瞳；昂贵，难以在基层推广普及

名称	时间	图片(张)	分辨率(像素)	数据集利用	拍摄(°)	任务	多专家标注
FP
DRIVE^[11]	2004	40	768×584	血管及EX	45	分割血管	是
DIARETDB0^[12]	2006	130	1500×1152	血管及EX	50	DR检测和分级	是
DiaRetDB1^[13]	2007	89	1500×1152	血管及EX	50	DR检测和分级	是
DRiDB^[14]	2013	50	768×584	MA，HE、EX，NVE	45	检测病灶，分割视盘、黄斑、血管	是
E－Ophtha^[15]	2013	463	多样	MA及EX	50	EX及MA检测	是
DR1^[16]	2014	1014	640×480	EX、HE和MA	45	DR检测	是
DR2^[17]	2014	520	867×575	所有特征	45	DR检测	－
Messidor^[18]	2014	1200	1440×960，2240×1488，2304×1536	HE及EX	45	DR及DME分级	是
KaggleEyePACS	2015	88702	多样	所有特征	－	DR分级	否
IDRiD^[19]	2018	516	4288×2848	MA，EX，HE	50	DR分级与病变分割	是
KaggleAPTOS 2019	2021	5590	多样	所有特征	－	DR分级	否
DDR^[20]	2019	12522	多样	所有特征	45	DR分级与病变分割	是
Messidor 2^[18]	NR	1748	多样	EX，HE	45	DR分级	是
STARE^[21]	2017	400	700×605	血管	35	视网膜血管分割	－
CHASE－DB1^[22]	2012	28(儿童)	1280×960	血管	30	视网膜血管分割	－
ROC^[23]	2009	100	768×576，1058×1061，1389×1383	MAs	45	HE和MA检测	是
ARIA^[24]	2017	143	768×576	MA、HE，EX	50	DR分级和血管分割	是
FFA
FFA PhotographsCF^[25]	2012	140	576×720	所有特征	－	DR分级和疾病检测	－
FFA Photographs^[26]	2014	70	576×720	MAs，血管	－	DR分级和病变检测	－
OCT
Rabbani	2015	24	－	－	－	糖尿病眼病	－
OCTID^[27]	2018	470	512×1024	黄斑水肿	－	DR分类	是
OCTAGON^[28]	2019	213	320×320	FAZ区域及血管分割	－	DR检测	是

Please choose a citation manager

Content to export