Visual impairment prevention by early detection of diabetic retinopathy based on stacked auto-encoder

Abstract

Diabetic retinopathy (DR) presents a significant concern among diabetic patients, often leading to vision impairment or blindness if left untreated. Traditional diagnosis methods are prone to human error, necessitating accurate alternatives. While various computer-aided systems have been developed to assist in DR detection, there remains a need for accurate and efficient methods to classify its stages. In this study, we propose a novel approach utilizing enhanced stacked auto-encoders for the detection and classification of DR stages. The classification is performed across one healthy (normal) stage and four DR stages: mild, moderate, severe, and proliferative. Unlike traditional CNN approaches, our method offers improved reliability by reducing time complexity, minimizing errors, and enhancing noise reduction. Leveraging a comprehensive dataset from KAGGLE containing 35,126 retinal fundus images representing one healthy (normal) stage and four DR stages, our proposed model demonstrates superior accuracy compared to existing deep learning algorithms. Data augmentation techniques address class imbalance, while SAEs facilitate accurate classification through layer-wise unsupervised pre-training and supervised fine-tuning. We evaluate our model's performance using rigorous quantitative measures, including accuracy, recall, precision, and F1-score, highlighting its effectiveness in early disease diagnosis and prevention of blindness. Experimental results across different training/testing ratios (50:50, 60:40, 70:30, and 75:25) showcase the model's robustness. The highest accuracy achieved during training was 93%, while testing accuracy reached 88% on a training/testing ratio of 75:25. Comparative analysis underscores the model's superiority over existing methods, positioning it as a promising tool for early-stage DR detection and blindness prevention.

Keywords: Deep learning; Diabetic retinopathy; Disability; Dropout; Stacked auto-encoder.

Fig. 1

Stages of Diabetic Retinopathy. Panel (a) shows a normal retina, while panels (b) to (d) illustrate NPDR stages. Panel (e) represents PDR.

Fig. 2

CNN architecture and steps involved in DR detection and stages.

Fig. 3

Stages of proposed methodology.

Fig. 4

Encoder 1.

Fig. 5

Encoder 2.

Fig. 6

Encoder 3.

Fig. 7

Encoder 4.

Fig. 8

Encoders are stacked.

Fig. 9

A neural network without dropout approach, all neurons are connected.

Fig. 10

A neural network with dropout approach, some neurons are disconnected.

Fig. 11

Developed framework of diabetic retinopathy classification.

Fig. 12

Accuracy of non-augmented data.

Fig. 13

Accuracy without dropout.

Fig. 14

Accuracy of augmented data and dropout.

Fig. 15

Accuracy at 50:50.

Fig. 16

Loss at 50:50.

Fig. 17

Performance metrics of a balanced dataset at 50:50.

Fig. 18

ROC curves of classes (0–4) at 50:50.

Fig. 19

Accuracy at 60:40.

Fig. 20

Loss at 60:40.

Fig. 21

Performance metrics of all classes (0–4) at 60:40.

Fig. 22

ROC curves of classes (0–4) at 60:40.

Fig. 23

Accuracy at 70:30.

Fig. 24

Loss at 70:30.

Fig. 25

Performance metrics of a balanced dataset at 70:30.

Fig. 26

ROC curves of classes (0–4) at 70:30.

Fig. 27

Accuracy at 75:25.

Fig. 28

Loss at 75:25.

Fig. 29

Performance metrics of all classes (0–4) at 75:25.

Fig. 30

ROC curves of classes (0–4) at 75:25.

Fig. 31

Predictions Made by the Model on the External Dataset for Each Class (Healthy, Mild, Moderate, Severe, Proliferative DR).

Fig. 32

Accuracy comparison with existing models.

Fig. 33

Sensitivity comparison with existing models.

Fig. 34

Specificity comparison with existing models.

Fig. 35

Precision comparison with existing models.

Fig. 36

F1-Score comparison with existing models.