Automatic Detection of Diabetic Retinopathy Based on the Presence of Micro-Aneurysms: Barriers to the Translation from Research to Clinical Practice

Diabetic retinopathy (DR) is a progressive disease which is the most common cause of blindness and early detection of this disease can limit vision loss. Retinal image grading is crucial for the prevention of visual loss but it is also a time consuming task. The Micro-aneurysm (MA) is the first sign of DR. The automatic detection of Micro-aneurysms (MAs) is an important and challenging research problem. This work presents an analysis of the failure modes of current automatic MA detection algorithms. It shows that such algorithms are confused by vascular structures. A method to improve algorithm performance by discarding vascular structures is proposed and analysed.