Computational framework to understand the clinical stages of COVID-19 and visualization of time course for various treatment strategies.

Coronavirus disease 2019 is known to be regulated by multiple factors such as delayed immune response, impaired T cell activation, and elevated levels of proinflammatory cytokines. Clinical management of the disease remains challenging due to interplay of various factors as drug candidates may elicit different responses depending on the staging of the disease. In this context, we propose a computational framework which provides insights into the interaction between viral infection and immune response in lung epithelial cells, with an aim of predicting optimal treatment strategies based on infection severity. First, we formulate the model for visualizing the nonlinear dynamics during the disease progression considering the role of T cells, macrophages and proinflammatory cytokines. Here, we show that the model is capable of emulating the dynamic and static data trends of viral load, T cell, macrophage levels, interleukin (IL)-6 and TNF-α levels. Second, we demonstrate the ability of the framework to capture the dynamics corresponding to mild, moderate, severe, and critical condition. Our result shows that, at late phase (>15 days), severity of disease is directly proportional to pro-inflammatory cytokine IL6 and tumor necrosis factor (TNF)-α levels and inversely proportional to the number of T cells. Finally, the simulation framework was used to assess the effect of drug administration time as well as efficacy of single or multiple drugs on patients. The major contribution of the proposed framework is to utilize the infection progression model for clinical management and administration of drugs inhibiting virus replication and cytokine levels as well as immunosuppressant drugs at various stages of the disease.

Continuous diagnosis and prognosis by controlling the update process of deep neural networks.

Continuous diagnosis and prognosis are essential for critical patients. They can provide more opportunities for timely treatment and rational allocation. Although deep-learning techniques have demonstrated superiority in many medical tasks, they frequently forget, overfit, and produce results too late when performing continuous diagnosis and prognosis. In this work, we summarize the four requirements; propose a concept, continuous classification of time series (CCTS); and design a training method for deep learning, restricted update strategy (RU). The RU outperforms all baselines and achieves average accuracies of 90%, 97%, and 85% on continuous sepsis prognosis, COVID-19 mortality prediction, and eight disease classifications, respectively. The RU can also endow deep learning with interpretability, exploring disease mechanisms through staging and biomarker discovery. We find four sepsis stages, three COVID-19 stages, and their respective biomarkers. Further, our approach is data and model agnostic. It can be applied to other diseases and even in other fields.