Development and Validation of an ICU Communication Checklist for Conscious Patients With Artificial Airways: Evidence Synthesis and Nominal Group Technique.

It is projected that by 2040, close to 600 million people will be living with diabetes with one-third of them likely to develop diabetic retinopathy (DR) as the most common cause of vision loss among working-age adults worldwide. DR is a neurovascular complication in which processes of AGE accumulation, PKC activation and VEGF upregulation participate thus degrading the inner blood-retinal barrier. Typical appearances are microaneurysms, intraretinal hemorrhages, hard and soft exudates and neovascular proliferation. Some forms of DR is presently experienced by approximately 90 million people and over 30 million have vision threatening conditions including proliferative DR (PDR) and diabetic macular edema (DME). The international diabetes federation estimates that diabetic patients will surpass 700 million by 2045 which will carry an unparalleled burden on healthcare systems especially in resource constrained environments. There is a lot of clinical importance in early diagnosis. The landmark DCCT and UKPDS trials revealed that HbA1c, when kept below 7%, decreases the risk of DR by up to 76% in type 1 diabetes and 25% in type 2 diabetes. Regardless of this, the global coverage of screening has been low because of labour shortage, geographic differences and other inter-observer variability, which highlights the necessity to develop scalable automated screening solutions. DR is analyzed according to international clinical diabetic retinopathy severity scale, which is based on Mild NPDR (microaneurysms only) and Moderate and Severe NPDR defined by the ETDRS 4-2-1 rule to PDR, where the neovascularization is necessitated by urgent treatment. Manual grading is subjective, time-consuming, and inaccurate, with the consensus of the experts being as low as 70–80% of cases and throughput mostly restricted to 100–120 patients daily which is a long way short of the needs of large-scale screening. The solution is provided by deep learning, especially Convolutional Neural Networks (CNNs). CNNs can be trained on large datasets with labeled labels that have proven to perform at the level of an expert in medical imaging, which has strongly suggested the use of these in automated screening and grading of DR. In the world, diabetic related retinopathy is the main cause of blindness particularly in countries that are resource deprived. This is the reason why a good and precise diagnostic system is highly required. Figure  demonstrates different phases of diabetic retinopathy.