Performance and limitation of machine learning algorithms for diabetic retinopathy screening and its application in health management: a meta-analysis

Moannaei, Mehrsa; Jadidian, Faezeh; Doustmohammadi, Tahereh; Kiapasha, Amir Mohammad; Bayani, Romina; Rahmani, Mohammadreza; Jahanbazy, Mohammad Reza; Sohrabivafa, Fereshteh; Asadi Anar, Mahsa; Magsudy, Amin; Sadat Rafiei, Seyyed Kiarash; khakpour, Yaser

doi:10.1186/s12938-025-01336-1

Table 2 Baseline characteristics of included studies and performance of color fundus image screening

From: Performance and limitation of machine learning algorithms for diabetic retinopathy screening and its application in health management: a meta-analysis

Code	Author (year) [Ref.]	Total images	Sensitivity	Specificity	AUC	Approach for screening
2	Jain (2021) [26]	1294	100	89.55		Kowa VX-10a mydriatic camera & Remidio FOP NM-10
3	Keel (2018) [27]	96	92.3	93.7	0.95	DLA
4	Jiang (2020) [28]	3228	93.9	94.4	0.94	Grad-CAM
5	Ipp (2020) [29]	893	95.5	85		ETDRS scale
6	Ibáñez-Bruron (2021) [30]	89	100	55.4		DART
7	Yo-Ping Huang (2020) [31]	52	96.6	95.2	0.99	VGG16, VGG19, MobileNet, InceptionV3, DenseNet
8	Hsu (2021) [32]	13,410	96.84	89.44	0.97	DLA
9	Yi-Ting Hsieh (2019) [33]	7524	92.2	97.5	0.95	CNN (VeriSee)
10	Heydon (2020) [34]	30,405	95.7	68		EyeArt v2.1
11	He (2019) [35]	889	90.79	98.5	0.94	Airdoc
12	Hao (2022) [36]	6146	79.2	87.1		VoxelCloud
13	Guo (2021) [37]	978	54	95	0.88	ResNet & U-Net
14	Gulshan (2019) [38]	3049	83.5	98.7	0.96	ICDR scale
15	Gulshan (2016) [39]	4997	90.3	98.1	0.99	DLA
16	Grzybowski (2021) [40]	60	93.33	94.45	0.94	Retinalyze
17	González-Briceño (2020) [41]	3368	89	92		Cross-industry standard process for data mining
18	Gargeya (2017) [42]	75,137	94	98	0.97	DLA
19	Glinton (2022) [43]	597	91	95	0.93	Python (version 3.6.9)
20	Gadekallu (2020) [44]	1151	90.4	94.3		DLA
21	Fleming (2023) [45]	179,944	89.19	77.41	0.99	DLA
22	M. Al-hazaimeh (2022) [46]	88,702	99.2	96.4	0.98	SVMGA
23	TamoorAziz (2023) [47]	219	94.21	97.46	0.98	DLA
24	Ghadah Alwakid (2023) [48]	12,522	89	99		CLAHE, ESRGAN
25	Eman AbdelMaksoud (2022) [49]	3662	96	69	0.99	CNN
26	Marc Baget-Bernaldiz (2021) [50]	1200	97.92	99.91	0.99	DLA
27	Anas Bilal (2022) [51]	98	96.9	96.9	0.97	U-NET, CNN-SVD
28	Usharani Bhimavarapu (2023) [52]	88,702	96.34	96.74	0.89	CNN
29	Wejdan L. Alyoubi (2021) [53]	13,673	89	97.3	0.95	CNN512, YOLOv3
30	Miao (2022) [54]	35,126	79.01	89.07	0.79	DLA
31	Penha (2023) [21]	686	93.6	71.7	0.86	EyerMaps
32	Lee (2021) [55]	311,604	80.47	81.28		VA HCS
33	Lam (2018) [56]	1346	95	96		DIGITS
34	Nugroho (2021) [57]	200	95	81		DLA
35	Nneji (2022) [58]	35,126	98.9	98	0.99	WFDLN
36	Zhang (2022) [22]	1089	98.23	74.45	0.95	EyeWisdom V1
37	Yang (2021) [59]	1418	79.6	79.9	0.81	XGBoost, RF, naïve Bayes, KNN, AdaBoost, Light GBM, ANN, LR
38	Zhao (2022) [60]	7943	88.9	74	0.8	RF, XGBoost, LR, SVM, KNN
39	Pinedo-Diaz (2022) [61]	420	97.66	98.33	0.98	DLA
40	Surya (2023) [62]	1085	83.33	98.86	0.83	Dr Noon
41	Zhang (2020) [63]	47,269	83.3	92.5		DLA
42	Sosale (2020) [64]	922	93	92.5	0.9	Medios AI
43	Mehboob (2022) [65]	25,600	78	44		DLA
44	Mujeeb Rahman (2022) [66]	560	93.65	95.13	0.97	DNN, SVM
45	Abramoff (2016) [67]	1748	96.7	87	0.98	IDx-DR X2.1
46	Palaniswamy (2023) [68]	813	94.28	99.34	0.96	DLA
47	Ting (2017) [69]	71,896	90.5	91.6	0.93	DLA
48	Jebaseeli (2019) [70]	201	80.61	99.54		DLBSVM
49	Jena (2022) [71]	100	99.2	99.4	0.99	2-branch CNN
50	Jiang (2019) [72]	30,244	85.57	90.85	0.946	DLA
51	Khan (2023) [73]	45	79.63	98.63	0.98	Inception v3 & DenseNet-121
52	Shankar (2020) [74]	541	98.54	99.38		SDL
53	Kuna (2023) [75]	1200	98.9	99.7		DL-DRDC
54	Ludwig (2020) [76]	92,364	89	89	0.89	CNN
55	Sosale (2020) [77]	297	98.84	86.73	0.92	ICDRS scale
56	Li (2022) [23]	1674	95	85.1	0.94	Deep learning algorithm
57	Roy (2020) [78]	1330	94	95		DLA
58	Romero-Aroca (2020) [79]	1748	96.7	97.6		DLA
59	Pei (2022) [80]	324	91	81.3	0.86	EyeWisdom
60	Rayave (2022) [81]	650	65.54	100		CNN, RNN, SVM, FKM, DA
61	Paradisa (2020) [82]	89	99.3	98		CNN, SVM, KNN, RF, XGBoost
62	Li (2022) [83]	950	97.96	93.88	0.99	NNs, SVM, XGBoost, DT, LR, GNB, KNN
63	Roychowdhury (2013) [84]	1200	100	53.16	0.87	GMM, SVM, KNN, AdaBoost
64	Sarao (2020) [85]	165	90.8	75.3	0.07	EyeArt
65	Li (2021) [86]	32,452	70	90	0.9	LR, XGBoost, RF, SVM
66	Wu (2022) [87]	7033	100	37.8	0.9	OC-Net, SE-Net
67	Ruamviboonsuk (2022) [88]	138	91.4	95.4		DLA
68	Ruamviboonsuk (2019) [89]	25,326	97	96		DLA
69	Saxena (2020) [90]	56,839	81.02	86.09	0.92	CNN
70	Sayres (2019) [91]	1612	79.4	96.6		DLA
71	A. Shah (2021) [92]	2680	100	81.82	0.98	IDx-DR
72	P. Shah (2020) [93]	1533	99.7	98.5	0.99	CNN
73	Rajalakshmi (2018) [94]	296	95.8	80.2		ICDR
74	Reddy (2022) [95]	89	90.2	95.2	0.88	DLA
75	Rom (2022) [96]	363	45	94	0.81	CNN
76	Rogers (2021) [97]	22,180	81.6	81.7	0.98	Pegasus
77	Ryu (2022) [98]	918	67.5	94.4		CNN

DLA: Deep-learning algorithm; VA: Veterans Affairs; HCS: Puget Sound Health Care System; DIGITS: Deep Learning GPU Training System; WFDLN: weighted fusion deep learning network; XGBoost: extreme gradient boosting; KNN: k-nearest neighbour; ANN: artificial neural network; RF: Random Forest; LR: Logistic Regression; SVM: Support Vector Machine; DNN: deep neural network; DLBSVM: Deep Learning Based Support Vector Machine; SDL: Synergic deep learning; CNN: convolutional neural networks; ICDRS: International Clinical Diabetic Retinopathy Severity Scale System; RNN: Recurrent Neural Network; FKM: Fuzzy K-means cluster; DA: Discriminant Analysis; GNB: Gaussian Naive Bayes; LR: Logistics Regression; DT: Decision Tree; NNs: Neural Networks; GMM: Gaussian Mixture model; OC-Net: occurrence network; SE-Net: a severity network; DL-DRDC: w deep learning empowered diabetic retinopathy detection and classification

Back to article page

ISSN: 1475-925X

Contact us

Submission enquiries: journalsubmissions@springernature.com

BioMedical Engineering OnLine

Contact us