Computer-Aided Retinal Haemorrhage Detection and Super-Resolution in Diabetic Retinopathy Digital Fundus Images

Retinal diseases such as diabetic retinopathy cause blood vessel ruptures that can lead to vision loss. Automatic retinal haemorrhage detection and labelling support clinicians to facilitate fast screening and initiate immediate treatment. This work proposes a haemorrhage localization technique for...

Full description

Saved in:
Bibliographic Details
Published inTENCON 2021 - 2021 IEEE Region 10 Conference (TENCON) pp. 209 - 214
Main Authors Salam, Amritha Abdul, Pal, Poulomi, Mahadevappa, Manjunatha
Format Conference Proceeding
LanguageEnglish
Published IEEE 07.12.2021
Subjects
Box-regression
Computational modeling
Haemorrhage Detection
Location awareness
Retinopathy
Shape
Super-Resolution
Superresolution
Telemedicine
Transfer learning
Online AccessGet full text

Cover

More Information
Summary:Retinal diseases such as diabetic retinopathy cause blood vessel ruptures that can lead to vision loss. Automatic retinal haemorrhage detection and labelling support clinicians to facilitate fast screening and initiate immediate treatment. This work proposes a haemorrhage localization technique for the accurate detection of retinal haemorrhages using transfer-learning. The retinal haemorrhages are localized using bounding boxes to obtain their corresponding coordinate locations within the fundus image. An attempt was made to develop a model for the detection of retinal haemorrhages using SSD architecture, for fundus images and the performance of the algorithm was evaluated on Diabetic Retinopathy databases available online. Moreover, a comparison on the haemorrhage detection is assessed by employing a transfer learning-based super-resolution technique on low-resolution images for a scale factor of 4. The proposed model gives a detection accuracy of 91.01%.
ISSN:2159-3450
DOI:10.1109/TENCON54134.2021.9707389