Integration of a procedure for automatic computation and monitoring of prognostic variables in retinographies in health information systems

Diabetic retinopathy (DR) is a complication of diabetes mellitus (DM) that causes loss of vision and can lead to blindness. The DR is expressed in the retina by means of four types of injury, that can be seen in eye fundus or retinography images: microaneurysms, microbleeds, hard exudates, and cotton wool spots. Based on the ongoing presence of these lesions, it is possible to establish the following gradation in the disease: mild, moderate and severe nonproliferative DR, and proliferative DR.
In the early stages of RD, the patient experiences no appreciable change in their vision so that in many cases, the presence of retinopathy overlooked. It is in these early stages when treatment to slow progression of the disease would be more effective. For this reason, one of the main priorities in prevention recommended by WHO are monitoring programs (screening) of patients. The detection and monitoring of the RD injuries would prevent loss of vision in the patient as well as cost savings to the public health.
For this reason, this project is the development of a methodology and monitoring automatic detection of microaneurysms in early stages of DR. The ultimate goal is the integration of these methodologies in health information systems both at the hospital and the primary care centers.

Objectives

  1. Creating an algorithm for the detection of red dots on retinographies: We depart from algorithms for detecting red spots proposed in previous works of the research team and an adjustment to suit the characteristics of the images obtained in health centres will be performed. The existence of an automatic algorithm will process instantly fundus images, processed manually which would take several minutes to a specialist.
  2. Development of a procedure for monitoring the evolution of the red dots (aneurysms) in patients: develop an algorithm based on registration techniques to align fundus images and trace the evolution of microaneurysms in a patient in time. Metrics to establish corresponding points between two images of the same patient will be defined.
  3. Clinical validation of the proposed algorithms: Several methods will be designed to clinically validate the algorithm for detecting red spots on different sets of retinographies. The results will be analyzed from a statistical standpoint.
  4. Integration of process in health information systems: Once validated the extraction methodology and monitoring, it will take place the integration of previously developed procedures in health information systems. Thus, it will be possible to carry out studies on the incidence and evolution of the DR in the Galician population.