Paper Info
Title
Computerized Diagnosis of Fundus Vascular Structure Based on Predictions of Diabetic Retinopathy Grade and Risk of Macular Edema
Abstract
Many diseases have been identified on the basis of auxiliary diagnosis from fundus images. The advancement of medical image analysis, some researchers can even predict the risk that individuals will suffer from some diseases in the future. Additionally, fundus vascular structure, which is a microvasculature, can be directly observed and plays an indispensable role in clinical research. Therefore, understanding how valuable are the vascular characteristics in computer-aided diagnosis is undeniably meaningful. Thus, we will analyze the importance of fundus vascular by predicting the retinopathy grade and risk of macular edema base on fundus images. Firstly, experiments are designed to isolate the existing neural network models, and then ensemble these models for superior performance. Finally, the ultimate model is applied onto analyzing the fundus vascular and the skeleton vascular respectively for reflecting the importance of vascular features. The F-2 scores of predicting the diabetic retinopathy grade on the entire fundus image, fundus vascular and skeleton vascular are 93.7%, 92.4% and 78.7% respectively. In predicting the risk of macular edema, the results are 96.8%, 93.3% and 81.0% respectively. From this perspective, vascular feature plays an indispensable role in computer diagnosis and the complete vascular information is beneficial for the improvement of the diagnostic effect.
Year
DOI
Venue
2019
10.1166/jmihi.2019.2688
JOURNAL OF MEDICAL IMAGING AND HEALTH INFORMATICS
Keywords
DocType
Volume
Fungus Images,Macular Edema,Computer-Aided Diagnosis,Neural Network
Journal
9
Issue
ISSN
Citations 
5
2156-7018
0
PageRank 
References 
Authors
0.34
0
4
Name
Order
Citations
PageRank
Zhi Liu100.34
Zhaomin Yao200.34
Yankun Cao334.74
Jianhuang Wu46011.75