DIABETIC RETINOPATHY DIAGNOSIS USING A CONVOLUTIONAL NEURAL NETWORK
| Autor: | Daria S. Shishikina, Vladimir I. Gorbachenko, Mikhail A. Shcherbakov |
|---|---|
| Jazyk: | English<br />Russian |
| Rok vydání: | 2023 |
| Předmět: | |
| Zdroj: | Модели, системы, сети в экономике, технике, природе и обществе, Iss 4 (2023) |
| Druh dokumentu: | article |
| ISSN: | 2227-8486 90149246 |
| DOI: | 10.21685/2227-8486-2023-4-7 |
| Popis: | Background. The paper deals with the problem of diagnosing diabetic retinopathy. The aim of the work is to create software to facilitate the diagnosis of diabetic retinopathy. Materials and methods. A publicly available set of ocular fundus images from Kaggle website was used as initial data for diabetic retinopathy diagnosis. Inception v3 convolutional neural network was used to work with these images. Quality metrics (precision, completeness and F1-measure) were calculated to assess the quality of the network. Results. Training of a convolutional neural network on ocular fundus images with signs of diabetic retinopathy for image-based diagnosis of the disease and its degree of development was carried out. The network was tested and its quality was evaluated. Conclusions. The network classifies images according to the generally accepted classification of diabetic retinopathy depending on the degree of its development. The network has the ability to classify and can be finalized experimentally for the possibility of its further implementation in professional medical activity. |
| Databáze: | Directory of Open Access Journals |
| Externí odkaz: |
|