Advances in retinal microaneurysms detection, segmentation and datasets for the diagnosis of diabetic retinopathy: a systematic literature review

Microaneurysms (MAs) are small, circular, red lesions typically observed in the initial stage of Diabetic Retinopathy (DR). DR is a degenerative eye condition resulting from diabetes mellitus that can lead to vision loss if not detected and treated early. Since MAs are one of the first visible signs of DR, detecting them can lead to early diagnosis and treatment of DR and potentially prevent severe vision loss. In this survey, we conduct a thorough review and analyze the published literature on automated detection and segmentation of MAs, with the aim of bolstering ophthalmologists capabilities in the early screening and management of DR. Section 2 provides an overview of publicly available datasets that include image-level and pixel-level annotations of MAs. Benchmarking plays a vital role in the precise assessment of the efficacy of early DR detection systems. Section 3 presents common metrics used to benchmark the performance of MA recognition technologies. We categorize MA recognition methodologies into two principal groups: those based on image processing and those leveraging deep learning techniques. Sections 4 and  5 meticulously enumerate the cutting-edge techniques for MA detection and segmentation. Our discussion also delves into recent advancements and explores prospective future directions in this field. In conclusion, the evolution of machine learning and computer vision has enabled automated MA recognition methods to exhibit considerable potential, marking a significant stride forward in the fight against DR-related vision loss.