Abstract
Purpose :
Diabetic Macular Edema is one of the main causes of legal blindness in work-age population. The Artificial Neural Networks are useful for early detection of Diabetic Macular Edema, with a great importance in developing countries, where specialized management is limited. The purpuse of the study was to determine the sensitivity and specificity in diagnostic test based on Artificial Neural Networks as a method for automatic detection of clinical features on Diabetic Macular Edema diagnosis
Methods :
A cross-sectional diagnostic test study for an Artificial Neural Networks based learning system was performed. Sensitivity and specificity for Diabetic Macular Edema diagnosis and classification of color fundus photographs were assessed
Results :
The Network shown an accuracy rate of 73.5% in detecting Diabetic Macular Edema during the training stage. In the final stage, for Diabetic Macular Edema diagnosis the sensitivity was 61% and specificity 69% (positive predictive value 63% and negative predictive value 67%) when compared with Gold Standard. To classify the grade of Diabetic Macular Edema the Network has a sensitivity of 70% and specificity 61% (with a positive predictive value 64% and negative predictive value 68%)
Conclusions :
Artificial Neural Networks show good performance in detection and classification of diabetic macular edema. This study is the first step to build a telemedicine tool to support physicians to detect Diabetic Macular Edema using color fundus photographs
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.