Artificial Intelligence for the Diagnosis of De Novo Diabetic Retinopathy: A Scoping Review of Global Evidence

Introduction: Diabetic retinopathy (DR) requires early diagnosis, as it is a common complication of diabetes mellitus (DM) and a leading cause of blindness worldwide. Artificial intelligence (AI) is emerging as a promising tool for detecting DR, as it can analyse large volumes of data with high accuracy (ACC). Methodology: A scoping review of the literature from 2019–2025 was conducted in four databases, and inclusion and exclusion criteria were applied for the selection of articles. Data extraction was performed, and a synthesis of the results was generated. Results Sixty-two articles were reviewed, mostly experimental studies (43%) and deep learning (DL) models (82%). AI was mainly applied to retinography (59%) and other images, such as optical coherence tomography (OCT) and optical coherence tomography angiography (OCTA) (11.3%). It demonstrated high sensitivity (SE) and specificity (SP) and highlighted benefits such as mass screening capacity. Limitations, such as image quality and variability between devices, have also been identified, but there are future opportunities to revolutionise the diagnosis of DR. Conclusion AI is emerging as an effective tool for the de novo diagnosis of DR, with retinography as the main technique and other methods as options with great potential. Its benefits include automation, greater coverage, and a reduced healthcare burden, reinforcing its diagnostic value. However, some technical and validation challenges need to be overcome before these methods can be implemented in clinical practice.