Grad-CAM for Visualizing Diabetic Retinopathy
Authors:
- 27 May 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Abstract
Diabetic Retinopathy (DR) is damage to the retina that occurs due to high blood glucose. This disease can lead to early blindness and hence it is important to diagnose it at an earlier stage. The use of artificial intelligence to aid a doctor in identifying such disorders is beneficial as it reduces the time spent by them. State of art model using the convolutional neural network (CNN) has an accuracy of 75%, however, CNNs are black-box models, and it is difficult to interpret their functioning. Hence, the outcome provided by the model cannot be trusted which warrants the use of explainable AI techniques to provide explanations that justify the model’s output. In this paper, we have enhanced the state of art model using a CNN to identify DR which gave an accuracy of 92%. Gradient-Weighted Class Activation Map (Grad-CAM) explainability is used to verify the regions considered significant by the model. The application presents the classification outcome along with explainability thereby aiding the medical practitioners to arrive at a decision.Keywords
This publication has 9 references indexed in Scilit:
- Transfer Learning approach for grading of Diabetic RetinopathyJournal of Physics: Conference Series, 2021
- Pre-diagnosis of Diabetic Retinopathy using Blob Detection2020
- Diabetes detection using deep learning algorithmsICT Express, 2018
- Image Classification Using an Ensemble-Based Deep CNN2018
- Automated detection of diabetes using CNN and CNN-LSTM network and heart rate signalsProcedia Computer Science, 2018
- Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization2017
- A review on deep convolutional neural networks2017
- A Review on Recent Developments for Detection of Diabetic RetinopathyScientifica, 2016
- Convolutional Neural Networks for Diabetic RetinopathyProcedia Computer Science, 2016