Table 2 Comparative analysis of DR grading using proposed Model.

From: Multi-task deep learning framework combining CNN: vision transformers and PSO for accurate diabetic retinopathy diagnosis and lesion localization

 

Technique

Dataset

Accuracy (%)

Precision (%)

Recall (%)

F1-Score

Laily (2024)21

Sobel Segmentation + Linear SVC

Kaggle

44.34

48.26

44.34

41.76

Fousiya et al. (2022)22

Attention U-Net

Kaggle, Aptos

95.6

94.8

91.8

92.4

Li et al. (2022)23

DACNN

EyePACS, Messidor

93.8

94.6

90.7

92.6

Costaner et al. (2024)24

LBP + Wavelet Transform + SVM

Customized Dataset

95.59

96

97.96

96.97

Gayathri et al. (2021)25

Multipath CNN + J48 Classifier

Kaggle, Messidor, IDRiD

99.62

98

99.4

98.9

Chilukoti et al. (2024)26

Ensemble Transfer Learning

EyePACS, Messidor

90

92

92.4

93

Yadav et al. (2022)27

InceptionResNetV2

Kaggle

95

95.2

94

96

Ali et al. (2023)28

ResNet 50 A

APTOS 2019

98.3

98.4

98.4

98.5

Jacoba et al. (2023)29

AutoML for handheld retinal images

APTOS

97

97

97

96

Talukder et al. (2023)30

DenseNetEnsemble 121

Customized Dataset

100

100

100

100

Proposed MTN (CNN + ViT + PSO Fusion)

DRTiD

98.9

98.8

98.7

98.8

Back to article page