Title: Designing a learning model for mobile vision to detect diabetic retinopathy based on the improvement of MobileNetV2

Authors: Hieu Nguyen; Vinh P. Tran; Vuong T. Pham; Hien D. Nguyen

Addresses: University of Information Technology, Quarter 6, Linh Trung Ward, Thu Duc District, Ho Chi Minh City, Vietnam; Vietnam National University, Quarter 6, Linh Trung Ward, Thu Duc District, Ho Chi Minh City, Vietnam ' University of Information Technology, Quarter 6, Linh Trung Ward, Thu Duc District, Ho Chi Minh City, Vietnam; Vietnam National University, Quarter 6, Linh Trung Ward, Thu Duc District, Ho Chi Minh City, Vietnam ' Sai Gon University, 273 An Duong Vuong Street, Ward 3, District 5, Ho Chi Minh City, Vietnam ' University of Information Technology, Quarter 6, Linh Trung Ward, Thu Duc District, Ho Chi Minh City, Vietnam; Vietnam National University, Quarter 6, Linh Trung Ward, Thu Duc District, Ho Chi Minh City, Vietnam

Abstract: Diabetes is a leading cause for cases of blindness in the world. Early detection of diabetic retinopathy can prevent or delay diabetes-related blindness. Nowadays, a mobile phone is a necessary device for almost all people. It is usually used for some daily works. Thus, an application on a mobile phone for checking the eyes in order to diagnose diabetic retinopathy early is very helpful. In this article, a method for building the mobile application to detect diabetic retinopathy is studied based on MobileNetV2. This method is built on the improvements of depthwise separable convolution, combining the layers of linear bottleneck and inverted residuals. Those layers are not only effective to keep more useful features by using linear layers, but also do not increase the cost of computing a lot. The method was made more effective when it runs on mobile. The proposed method had been tested on two datasets EyePACs 2015 and APTOS 2019. The positive results of the experiment are emerging to build an application for detecting diabetic retinopathy which is used on mobile phones.

Keywords: diabetic retinopathy; diabetes; deep learning; ordinal regression; MobileNetV2; convolutional neural network; CNN; mobile application; software engineering.

DOI: 10.1504/IJDET.2022.124987

International Journal of Digital Enterprise Technology, 2022 Vol.2 No.1, pp.38 - 53

Received: 13 May 2020
Accepted: 29 Aug 2020
Published online: 22 Aug 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article