Diabetic Retinopathy Detection Using Transfer Learning and Deep Learning
Diabetic retinopathy is one of the major causes of blindness in the population aged 20–65. In this paper, we address the problem of automatic diabetic retinopathy detection and proposed a novel deep learning hybrid to solve the problem. We use transfer le
- PDF / 468,565 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 7 Downloads / 262 Views
Abstract Diabetic retinopathy is one of the major causes of blindness in the population aged 20–65. In this paper, we address the problem of automatic diabetic retinopathy detection and proposed a novel deep learning hybrid to solve the problem. We use transfer learning on pre-trained Inception-ResNet-v2 and added a custom block of CNN layers on top of Inception-ResNet-v2 for building the hybrid model. We evaluated the performance of the proposed model on Messidor-1 diabetic retinopathy dataset and APTOS 2019 blindness detection (Kaggle dataset). Our model performed better than other published results. We achieved a test accuracy of 72.33% and 82.18% on Messidor-1 and APTOS dataset, respectively. Keywords Diabetic retinopathy · Image classification · Deep learning · Inception ReseNet-v2
1 Introduction Diabetic retinopathy is one of the major causes of the rise in global blindness. As per records, there are as many as 415 million diabetic patients worldwide. To prevent blindness, diabetics should be screened every year. A common practice for detecting diabetic eye disease is to examine the fundus image and assess the severity of
A. K. Gangwar · V. Ravi (B) Center of Excellence in Analytics, Institute for Development and Research in Banking Technology, Castle Hills Road 1, Masab Tank, Hyderabad 500057, India e-mail: [email protected] A. K. Gangwar e-mail: [email protected] A. K. Gangwar School of Computer and Information Sciences, University of Hyderabad,Hyderabad 500046, India © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 V. Bhateja et al. (eds.), Evolution in Computational Intelligence, Advances in Intelligent Systems and Computing 1176, https://doi.org/10.1007/978-981-15-5788-0_64
679
680
A. K. Gangwar and V. Ravi
the disease. The severity depends on the type of retinopathy (e.g., microaneurysms, hemorrhages, hard exudates, etc.), which are indicators of eyeball bleeding and exudation. Despite having nearly 127,000 ophthalmologists in India, nearly 45% of patients suffer from blindness before being diagnosed . This should not happen because diabetic retinopathy is completely preventable. To improve this situation, the Google research team worked closely with EyePACS in the USA and three eye hospitals in India [1], Aravind Eye Hospital, Sankara Nethralaya, and Narayana Nethralaya. Therefore, building better deep learning-enabled software for automatic diabetic retinopathy detection becomes necessary. There is an increasing progress in collaboration between a diabetes care physician and an ophthalmologist. There are a few opportunities for diabetic patients to undergo fundus examinations in areas with few doctors. Further, there is a shortage of ophthalmologists who can diagnose and treat diabetic retinopathy. However, the level of dispensary ophthalmologic monitoring of patients with diabetes today remains unsatisfactory even in economically highly developed countries. For example, in the USA, about a third of diabetics ha
Data Loading...
