You are on page 1of 13

Eye Disease Classification

[ Mini Project - II ]
B. Tech. VI Semester , Department of Information Technology

Group Members : Instructor :-


Arihant Mishra -IEC2020032
Dr. Sanjai Singh
Rohit kumar Meena -IEC2020076
Satya Prakash Kumawat -IEC2020023
1. Problem Identification and Definition
2. Dataset Description
3. Literature Review
4. Methodology : Flowchart
TABLE OF
5. DC GAN Architecture
CONTENTS 6. Parameters Used in DCGAN
7. Synthetic Images
8. Augmented Data Analysis
9. Classification Using CNN
10. Result
11. Conclusion
12. References
Problem Identification

Develop a Model that can accurately Classify the Severity of Diabetic Retinopathy in
patients using Retinal Images to aid in early detection and treatment.

What is Diabetic Retinopathy?


Diabetic retinopathy is a diabetes complication that affects eyes. It's caused by
damage to the blood vessels of the light-sensitive tissue at the back of the eye
(retina).
DATASET DESCRIPTION
❖ Dataset used : APTOS 2019 BLINDNESS DATASET
❖ The dataset contains 4,996 retinal images, which were graded by ophthalmologists on a scale of
0 to 4 based on the severity of diabetic retinopathy.
❖ Number of Training Images : 3662
❖ Number of Testing Images : 1334

STAGES OF DIABETIC RETINOPATHY

Normal DR Mild DR Moderate DR Severe DR Proliferative DR


[Label 0] [Label 1] [Label 2] [Label 3] [Label 4]
LITERATURE REVIEW
Paper Title Authors Year Dataset Methodology Results

Comparing Deep Feature Extraction Nilarun 2023 Eye-PACS, Xception, 91%,


Strategies for Diabetic Retinopathy Mukherjee APTOS, ShuffleNet-V2 81.51%
Stage Classification from Fundus et al. Messidor-1,
Images. [1] Messidor-2

Applying supervised contrastive Md Robiul 2022 Messidor-1, Supervised 84.36%


learning for the detection of diabetic Islam et al. Messidor-2, Contrastive
retinopathy and its severity levels from EyePACS Learning
fundus images. [2]

Diabetic Retinopathy Grading by a Chenrui 2022 APTOS Source-Free 91.2%


Source-Free Transfer Learning Zhang et 2019 Transfer
Approach. [3] al. Learning
(SFTL)
METHODOLOGY : FLOWCHART
DC GAN Architecture
Label Original Images Synthetic Images

4
Classification Using CNN
● For the classification purpose we have experimented with several pre-trained CNN
models.
● Models evaluated:
■ ResNet50
■ DenseNet121
■ Efficient-Net B3
● We evaluated the performance of each CNN model both with and without attention
layers (spatial and channel attention) for image classification.
● In total, we assessed the performance of 6 CNN models with and without attention
layers
Image preprocessing
Crop: Focusing on Relevant Content
Cropping an image involves selecting a specific region of interest and discarding irrelevant or distracting elements.

Gaussian Blur: Smoothing and Noise Reduction


Gaussian blur is a widely used technique for reducing noise and smoothing an image we can minimize high-frequency
variations and random pixel fluctuations, resulting in a cleaner and more visually pleasing image.

Resize: Standardizing Image Dimensions:


Resizing an image involves adjusting its dimensions to a standardized size while preserving its aspect ratio also helps to
optimize computational efficiency, reducing processing time and memory requirements.
REFERENCES
[1] N. Mukherjee and S. Sengupta, "Comparing Deep Feature Extraction Strategies for Diabetic Retinopathy Stage Classification from
Fundus Images," in IEEE Arabian Journal of Science and Engineering, 2023. doi: 10.1007/s13369-022-07547-1.

[2] M. R. Islam et al., "Applying supervised contrastive learning for the detection of diabetic retinopathy and its severity levels from
fundus images," Computers in Biology and Medicine, vol. 146, pp. 1-11, Jul. 2022. doi: 10.1016/j.compbiomed.2022.105602.

[3] C. Zhang, T. Lei, and P. Chen, "Diabetic retinopathy grading by a source-free transfer learning approach," Biomedical Signal
Processing and Control, vol. 73, p. 103423, Feb. 2022. doi: 10.1016/j.bspc.2021.103423.

[4] J. D. Bodapati, "Stacked convolutional auto-encoder representations with spatial attention for efficient diabetic retinopathy
diagnosis," Multimedia Tools and Applications, vol. 81, no. 22, pp. 32033-32056, Sep. 2022. doi: 10.1007/s11042-022-12811-5.

[5] G. Yue, Y. Li, T. Zhou, X. Zhou, Y. Liu, and T. Wang, "Attention-Driven Cascaded Network for Diabetic Retinopathy Grading from
Fundus Images," Biomedical Signal Processing and Control, vol. 80, no. 2, pp. 104370, Feb. 2023. doi: 10.1016/j.bspc.2022.104370.

[6] P. Islam, M. Z. Islam, and M. A. Hossain, "An efficient framework for detecting and classifying diabetic retinopathy from fundus
images," Algorithms, vol. 13, no. 12, pp. 319, Dec. 2020. doi: 10.3390/a13120319.
[7] A. Sugeno, Y. Ishikawa, T. Ohshima, and R. Muramatsu, "Simple methods for the lesion detection and severity grading of diabetic
retinopathy by image processing and transfer learning," Comput. Biol. Med., vol. 132, p. 104795, Jul. 2021, doi:
10.1016/j.compbiomed.2021.104795.

[8] https://keras.io/examples/generative/

You might also like