A Hybrid Convolutional Neural Network Model for Automatic Diabetic Retinopathy Classification From Fundus Images

Abstract

Objective: Diabetic Retinopathy (DR) is a retinal disease that can cause damage to blood vessels in the eye, that is the major cause of impaired vision or blindness, if not treated early. Manual detection of diabetic retinopathy is time-consuming and prone to human error due to the complex structure of the eye. Methods & Results: various automatic techniques have been proposed to detect diabetic retinopathy from fundus images. However, these techniques are limited in their ability to capture the complex features underlying diabetic retinopathy, particularly in the early stages. In this study, we propose a novel approach to detect diabetic retinopathy using a convolutional neural network (CNN) model. The proposed model extracts features using two different deep learning (DL) models, Resnet50 and Inceptionv3, and concatenates them before feeding them into the CNN for classification. The proposed model is evaluated on a publicly available dataset of fundus images. The experimental results demonstrate that the proposed CNN model achieves higher accuracy, sensitivity, specificity, precision, and f1 score compared to state-of-the-art methods, with respective scores of 96.85%, 99.28%, 98.92%, 96.46%, and 98.65%.INDEX TERMS Diabetic retinopathy, fundus images, machine learning, computer vision. Clinical and Translational Impact Statement-Diabetic Retinopathy is becoming a more common cause of visual impairment in working-age individuals. Optimal results for preventing diabetic vision loss require patients to undergo extensive systemic care. Early detection and treatment are the key to preventing diabetic vision loss, which results from long-term diabetes and causes blood vessel fluid leakage of the retina. Common indicators of DR include blood vessels, exudate, hemorrhages, microaneurysms, and texture. To address this issue, this study proposes a novel CNN model for diabetic retinopathy detection. The proposed approach is an end-to-end mechanism that utilizes Inceptionv3 and Resnet50 for feature extraction of diabetic fundus images. The features extracted from both models are concatenated and input into the proposed InceptionV3 Resnet50 convolutional neural network (IR-CNN) model for retinopathy classification. To improve the performance of the proposed model, several experiments, including image enhancement and data augmentation methods, are conducted. The use of DL models, such as Resnet50 and Inceptionv3, for feature extraction enables the model to capture the complex features underlying diabetic retinopathy, leading to more accurate and reliable classification results. The proposed approach outperformed the existing models for DR detection.

Lyme rashes disease classification using deep feature fusion technique

“…64 The capabilities of DenseNet reduce the total number of parameters, and it has a minimal effect on the gradient problem, feature deployment, and feature reuse. 65 to change it; however, if the value is higher, it will not be changed and will continue to be used as is. At this stage, a ReLu-activated matrix image is multiplied by a convolution matrix with a 3 × 3 filter.…”

Segmentation and classification of skin lesions using hybrid deep learning method in the Internet of Medical Things

“…Wearable fitness trackers, remote patient monitoring systems, smart medical implants, and healthcare‐related smartphone apps all fall under this technology category. The IoMT has been a game‐changer in the medical field by supplying real‐time patient data and empowering doctors to provide better care and see better patient results 5–7 . This technology is essential in today's healthcare systems because of its ability to expand patient access to care, lower hospitalization rates, and increase the prevalence of preventative medicine through constant monitoring.…”

UC-stack: a deep learning computer automatic detection system for diabetic retinopathy classification

“…Deep learning models can achieve better performance compared with traditional methods by automatically optimizing features end-to-end, that is, directly inputting fundus images into the model to predict the grading results. Ali et al (2023) proposed a method to detect DR by using a convolutional neural network (CNN) model utilizing two different deep learning models Resnet50 and Inceptionv3 to extract features and feed them into the CNN for classification. The proposed method achieves high accuracy on OCT fundus image datasets.…”