EDR-Net: Lightweight Deep Neural Network Architecture for Detecting Referable Diabetic Retinopathy.

Present architecture of convolution neural network for diabetic retinopathy (DR-Net) is based on normal convolution (NC). It incurs high computational cost as NC uses a multiplicative weight that measures a combined correlation in both cross-channel and spatial dimension of layer's inputs. This might cause the overall DR-Net architecture to be over-parameterised and computationally inefficient. This paper proposes EDR-Net - a new end-to-end, DR-Net architecture with depth-wise separable convolution module. The EDR-Net architecture was trained with DRKaggle-train dataset (35,126 images), and tested on two datasets, i.e. DRKaggle-test (53,576 images) and Messidor-2 (1,748 images). Results showed that the proposed EDR-Net achieved predictive performance comparable with current state-of-the-arts in detecting referable diabetic retinopathy (rDR) from fundus images and outperformed other light weight architectures, with at least two times less computation cost. This makes it more amenable for mobile device based computer-assisted rDR screening applications.

Pediatric intensive care hybrid-style clinical round during COVID-19 pandemic: A pilot study

BackgroundWith the evolving COVID-19 pandemic and the emphasis on social distancing to decrease the spread of SARS-CoV-2 among healthcare workers (HCWs), our pediatric intensive care unit (PICU) piloted utilization of Zoom online into the clinical rounds to enhance communication among the treating team. We aimed to explore the feasibility of these hybrid virtual and physical clinical rounds for PICU patients from the HCWs perspective. MethodsA mixed quantitative and qualitative deductive thematic content analysis of narrative responses from pediatric intensive care HCWs were analyzed, descriptive statistics were used ResultsA total of 31 HCW were included in the analysis; the mean time of the virtual round was 72.45 minutes vs. 34.68 for physical rounds, the most shared component in the virtual round was CXR (93.5%). Some of the HCWs perceived advantages of the hybrid rounds were enabling the multidisciplinary discussions, lesser round interruptions, and practicality of the virtual discussions. The perceived challenges were the difficulty of the bedside nurse to attend the virtual round, decreased teaching opportunities for the trainees, and decreased interactions among the team members, especially if the video streaming was not utilized. ConclusionHybrid virtual and physical clinical rounds in PICU were perceived as feasible by HCWs. The virtual rounds decreased the physical contact between the HCWs, which could decrease the possibility of SARS-CoV-2 spread among the treating team. Still, several components of the hybrid round could be optimized to facilitate the virtual team-members interactions and enhance the teaching experience.