Analise da progressão da retinopatia diabética para identificar riscos e necessidade de tratamento

The identification of diabetic patients at the highest risk of DR progression and/or higher cardiovascular risk is an unmet need that needs to be addressed. The presence and severity of DR are associated with the risk of developing cardiovascular complications. The research work to be performed in this project will allow a better characterization of the pathophysiology of the DR and associated cardiovascular complications in real-world setting-based data. It will also allow the development of novel AI for DR staging and progression based on modern imaging techniques that include (ultra) widefield images ( and systemic metrics, and medication) as well as the estimation of the cardiovascular risk factors. Computing risk factors of cardiovascular diseases using AI models may lead to a significant reduction of the cost related to diabetic complications. After been identify in screening programs (or eventually in primary care) patients at risk of progression of DR or at risk of developing severe cardiovascular complications can be sent to more specialized care. The research work proposed in the project aims to address the following clinical questions: • Influence of the lesions that appear in the central field of view versus the ones that appear in the periphery in the pathways of the progression of DR • Association of the lesions/biomarkers that appear in the central field of view versus the ones that appear in the periphery on cardiovascular complications • Pathophysiology mechanisms associated with the progression of DR and the influence of systemic diseases like cardiovascular complications The objectives of this ambitious multidisciplinary project are: • Replacing ETDRS DRSS manual grading by automated methods based on retina images (OPTOS, CFP, OCT, and OCTA) and associated extracted metrics and tabular data (demographic, systemic, and medication) • Characterization of the association of major cardiovascular events with the retinal morphology and vascular network in the eyes of patients with T2D based on metrics extracted from images of the retina (OPTOS, CFP, OCT, and OCTA) and tabular data (demographic, systemic, and medication) • Development of models of DR progression based on metrics extracted from modern (ultra) widefield imaging and tabular data • Estimation of the risk factors associated with cardiovascular complications based on metrics extracted from images of the retina In this project, particular attention will be given to the interpretability, explainability and energy efficiency (training and inference) of the models. The massive adoption of AI has put pressure on the requirements that need to be followed during the development, validation, and deployment of these technologies. In critical areas such as healthcare, special care must be taken to assure robustness, transparency, interpretability, and explainability. In the scope of this project, the ALERT database will be created using data of approximately 700 patients. Is expected that the use of real-world setting data will allow the development of robust AI models and contribute to the proper characterization and validation of the models. The ALERT database will be developed following the FAIR principles and is expected, in the future, to be reused by other research groups.