Diabetic Retinopathy Detection From Fundus Images Using Multi-Tasking Model With EfficientNet B5

Abstract

Diabetic Retinopathy (DR) is a common eye disease that affects over 3 million people annually. People with diabetes are more prone to suffer from Diabetic Retinopathy. This condition can cause blurring of vision and blindness. Early detection and treatment are the most effective ways to manage Diabetic Retinopathy. Due to the huge number of diabetic patients and the need for more accurate and automatic diagnosis, the development of deep neural networks has been acknowledged. One of the issues with deep learning classification tasks, particularly in the medical field, is a lack of labelled training data. Transfer learning enables the deep learning model to be trained with a minimal Training Dataset. In Deep convolutional networks, transfer learning can be utilized to solve the problem of insufficient training data. Previous studies on deep neural networks have been promising. Through a color fundus photography study, we have developed a model to identify the different stages of Diabetic Retinopathy using Deep learning. Our proposed model has achieved 87% of accuracy with the help of EfficientNet model. The main aim of this work is to develop a robust system for detecting DR automatically.

Semantic segmentation of retinal exudates using a residual encoder–decoder architecture in diabetic retinopathy

“…The purpose of this research is to analyze and contrast the efficacy of various CNN techniques that are used in various research papers with different performance criteria. The most popular technique of transfer learning is used with the state‐of‐the‐art deep learning algorithms U‐Net (Ali et al, 2022), ResNet (Oh et al, 2021), VGG‐16 (Wang et al, 2018), Efficient Net (Bhawarkar et al, 2022), and Google Net (Li et al, 2019) for binary classification of exudates and non‐exudates using the MATLAB tool for semantic segmentation. All these networks are trained on the E‐ophtha database and then tested on the publicly available HEI‐MED database using performance metrics.…”

“…Deep learning was used in this research to create a model that can differentiate between the various phases of DR. With the assistance of the EfficientNet model, the accuracy of the suggested model has been improved to 87%. The primary objective of this effort is to create a reliable system that is capable of automatically identifying DR (Bhawarkar et al, 2022).Several deep learning-based automated systems with high sensitivity and specificity (>90%) for DR screening have been proposed. However, the current publicly accessible fundus imaging datasets are too small for these deep learning models to provide satisfactory results in clinical applications (Li et al, 2019).…”

“…FN: exudes pixels incorrectly as non-exuding pixels. et al, 2018), Efficient Net (Bhawarkar et al, 2022), and Google Net (Li et al, 2019) for binary classification of exudates and non-exudates using the MATLAB tool for semantic segmentation. All these networks are trained on the E-ophtha database and then tested on the publicly available HEI-MED database using performance metrics.…”

Nature inspired firefighter assistant by unmanned aerial vehicle (UAV) data

“…One of the most recent research projects as a firefighter assistant for a UAV subject belonging to (Ghali et al, 2022). They made an ensemble deep learning method by combining EfficientNet-B5 (Bhawarkar et al, 2022) and DenseNet-201 models in order to fire classification and segmentation tasks leading to 85 % accuracy on the FLAME dataset. Table 2 represents UAV-based related research on firefighter assistant subjects and for different tasks.…”

Development of preprocessing methods and revised EfficientNet for diabetic retinopathy detection

“…Recently, EfficientNet, a type of scaling model, has been used to simultaneously balance speed and accuracy and propose a more generalized idea for the optimization of the previous classification network. 16 Thus, many recent studies have adopted EfficientNet to develop a DR detection model [17][18][19] and suggested that EfficientNet is more efficient than many current wellperforming network architectures. In summary, the following are the drawbacks of traditional approaches and DL models: 1. limited dataset, 2. twisted and blurred images, 3. overfitting models, and 4. limited computing power.…”