COVID-MTL: Multitask learning with Shift3D and random-weighted loss for COVID-19 diagnosis and severity assessment.

There is an urgent need for automated methods to assist accurate and effective assessment of COVID-19. Radiology and nucleic acid test (NAT) are complementary COVID-19 diagnosis methods. In this paper, we present an end-to-end multitask learning (MTL) framework (COVID-MTL) that is capable of automated and simultaneous detection (against both radiology and NAT) and severity assessment of COVID-19. COVID-MTL learns different COVID-19 tasks in parallel through our novel random-weighted loss function, which assigns learning weights under Dirichlet distribution to prevent task dominance; our new 3D real-time augmentation algorithm (Shift3D) introduces space variances for 3D CNN components by shifting low-level feature representations of volumetric inputs in three dimensions; thereby, the MTL framework is able to accelerate convergence and improve joint learning performance compared to single-task models. By only using chest CT scans, COVID-MTL was trained on 930 CT scans and tested on separate 399 cases. COVID-MTL achieved AUCs of 0.939 and 0.846, and accuracies of 90.23% and 79.20% for detection of COVID-19 against radiology and NAT, respectively, which outperformed the state-of-the-art models. Meanwhile, COVID-MTL yielded AUC of 0.800 ± 0.020 and 0.813 ± 0.021 (with transfer learning) for classifying control/suspected, mild/regular, and severe/critically-ill cases. To decipher the recognition mechanism, we also identified high-throughput lung features that were significantly related (P < 0.001) to the positivity and severity of COVID-19.

Prediction of the Rehabilitation Duration and Risk Management for Mild-Moderate COVID-19.

OBJECTIVES: More than 80% of coronavirus disease 2019 (COVID-19) cases are mild or moderate. In this study, a risk model was developed for predicting rehabilitation duration (the time from hospital admission to discharge) of the mild-moderate COVID-19 cases and was used to conduct refined risk management for different risk populations. METHODS: A total of 90 consecutive patients with mild-moderate COVID-19 were enrolled. Large-scale datasets were extracted from clinical practices. Through the multivariable linear regression analysis, the model was based on significant risk factors and was developed for predicting the rehabilitation duration of mild-moderate cases of COVID-19. To assess the local epidemic situation, risk management was conducted by weighing the risk of populations at different risk. RESULTS: Ten risk factors from 44 high-dimensional clinical datasets were significantly correlated to rehabilitation duration (P < 0.05). Among these factors, 5 risk predictors were incorporated into a risk model. Individual rehabilitation durations were effectively calculated. Weighing the local epidemic situation, threshold probability was classified for low risk, intermediate risk, and high risk. Using this classification, risk management was based on a treatment flowchart tailored for clinical decision-making. CONCLUSIONS: The proposed novel model is a useful tool for individualized risk management of mild-moderate COVID-19 cases, and it may readily facilitate dynamic clinical decision-making for different risk populations.