A Large Language Modelling Deep Learning Framework for the Next Pandemic (preprint)

Deep neural networks have been integrated into the whole clinical decision procedure which can improve the efficiency of diagnosis and alleviate the heavy workload of physicians. Typical applications include 1) medical report generation, 2) disease classification, and 3) survival prediction. Since most neural networks are supervised, their quality heavily depends on the volume and quality of available labels. However, for novel diseases, e.g., new pandemics or new variants, there are few existing labels. In addition, the acquisition of new pandemic cases to collect sufficient labels for training is time-consuming and is typically unavailable at the early stage. To prepare neural networks for the next pandemic, in this paper, we propose a large language model - Unsupervised Learning from Unlabelled Medical Images and Text (ULUMIT) framework, which can learn broad medical knowledge (e.g., image understanding, text semantics, and clinical phenotypes) from unlabelled data. As a result, when encountering new pandemics, our framework can be rapidly deployed and easily adapted to them with extremely limited labels. Furthermore, ULUMIT supports medical data across visual modality (e.g., chest X-ray and CT) and textual modality (e.g., medical report and free-text clinical note); therefore, it can be used for any clinical task that involves both visual and textual medical data. We demonstrate the effectiveness of our ULUMIT by showing how it would perform using the COVID-19 pandemic ``in replay''. In particular, in the retrospective setting, we test the model on the early COVID-19 datasets; and in the prospective setting, we test the model on the new variant COVID-19-Omicron. The experiments are conducted on 1) three kinds of input medical data, image-only, text-only, and image-text; 2) three kinds of downstream tasks, medical reporting, diagnosis, and prognosis; 3) five public COVID-19 datasets; and 4) three different languages, i.e., English, Chinese, and Spanish. All experiments consistently show that our framework can make accurate and robust COVID-19 decision-support tasks with little labelled data (such as considering information from only one patient), providing an impact on medical data analysis during the early stage of the next pandemic. Besides COVID-19, our framework can be applied to identify 14 common thorax diseases and tuberculosis across five additional public datasets, demonstrating its robustness in generalization and transferability. In brief, our framework achieves state-of-the-art performances on ten datasets.

Global research trends in the COVID-19 and digestive disease: A review of visualization and bibliometric study

Background: The rapid spread of coronavirus disease 2019 (COVID-19) has attracted worldwide attention. There were also reported gastrointestinal symptoms in patients with COVID-19. This work aims to analyze the global research trends in COVID-19 and digestive disease. Methods: The related papers on COVID-19 and digestive disease were identified with Pubmed and web of science core collection on September 3, 2021. Bibliometric visualization was conducted through VOSviewer and CiteSpace. Results: The analytic research was based on original articles and reviews. There were 997 articles found, with citations ranging from 0 to 878. These articles were distributed among 86 countries and 355 journals. The USA mainly contributed (288 articles), where 3 of the top 10 institutions were located. Followed by China (215 articles) and Italy (160 articles). The highest level of scientific collaboration has been formed between the USA to China. The World Journal of Gastroenterology (39 papers) published the most significant number of articles. Concerning the research topic, the colon/small bowel had the largest number of articles, followed by the liver and pancreaticobiliary. "Liver injury,” "inflammatory bowel disease,” "management,” and "endoscopy” were the hotspot keywords. The largest cluster of liver transplantation had offered hints regarding research frontiers. Conclusion: The analytic results showed that the liver, especially liver transplantation, and inflammatory bowel disease were the 2 most influential research topics in COVID-19 and digestive disease.

Effect of methylprednisolone therapy on hospital stay and viral clearance in patients with moderate COVID-19

Background The benefits and harms of methylprednisolone treatment in patients with moderate coronavirus disease 2019 (COVID-19) remain controversial. In this study, we investigated the effect of methylprednisolone on mortality rate, viral clearance, and hospitalization stay in patients with moderate COVID-19. Methods This retrospective study included 4827 patients admitted to Wuhan Huoshenshan and Wuhan Guanggu hospitals from February to March 2020 diagnosed with COVID-19 pneumonia. The participants’ epidemiological and demographic data, comorbidities, laboratory test results, treatments, outcomes, and vital clinical time points were extracted from electronic medical records. The primary outcome was in-hospital death;secondary outcomes were time from admission to viral clearance and hospital stay. Univariate and multivariate logistic or linear regression analysis were used to assess the roles of methylprednisolone in different outcomes. The propensity score matching (PSM) method was used to control for confounding factors. Results A total of 1320 patients were included in this study, of whom 100 received methylprednisolone. Overall in-hospital mortality was 0.91% (12/1320);the 12 patients who died were all in the methylprednisolone group, though multivariate logistic regression analysis showed methylprednisolone treatment was not a risk factor for in-hospital death in moderate patients before or after adjustment for confounders by PSM. Methylprednisolone treatment was correlated with longer length from admission to viral clearance time and hospital stay before and after adjustment for confounders. Conclusions Methylprednisolone therapy was not associated with increased in-hospital mortality but with delayed viral clearance and extended hospital stay in moderate COVID-19 patients.

Machine learning elucidates the impact of short-term emission changes on air pollution in Beijing

The Chinese Spring Festival (CSF) is the most solemn traditional festival in China, and the substantial changes in anthropogenic activities in megacities provide a unique natural experiment to assess the influence of short-term emission changes on air quality. Here we applied a machine learning based random forest algorithm to six-year aerosol composition measurements in urban Beijing during the CSFs of 2012–2020 to quantify the relative contributions of meteorology and emission changes to air quality. Our results demonstrate large variabilities of air pollutants during the CSF due to the meteorological changes and holiday effect. By removing the meteorological effect, we found that the reduced emissions during CSF caused an average decrease of 5.1% for non-refractory PM2.5 with chloride and primary organic aerosol being the largest (8.8–18.7%) while the changes in secondary species were small. The COVID-19 lockdown during 2020 led to additional reductions of primary species by 16.3–36.8%, yet increases in nitrate and secondary organic aerosol due to enhanced secondary production. Our study has a significant implication that reducing local traffic and cooking emissions is far from enough for mitigating air pollution in winter in megacities due to the nonlinear effect of secondary production and regional transport. A synergetic control of multiple precursors, e.g., NOx and ammonia, is of great importance to reduce secondary aerosol and improve air quality.

Nitrate and secondary organic aerosol dominated particle light extinction in Beijing due to clean air action

Air quality in China has been continuously improved since clean air action in 2013, yet the visibility was not improved simultaneously. Here we employed a new method by integrating highly-time resolved aerosol compositions with particle light extinction (bext) into positive matrix factorization to quantify the different contributors to visibility degradation during four seasons in Beijing. Our results show that ammonium nitrate-related factor contributed dominantly to bext during all seasons (31–48%) and played more significant roles during low-visibility periods. Secondary organic aerosol (SOA) was an important contributor of bext (27–35%) in autumn and spring while primary OA related sources were more important in winter (37%). An increase in aerosol mass extinction efficiency and similarly important roles of ammonium nitrate and SOA in visibility degradation were also observed during COVID-19 lockdown. Our results point towards a future challenge in improving visibility in China due to the increased contributions of nitrate and SOA in PM2.5. Future emission controls with a priority to decrease nitrate would benefit both air quality and visibility.

A chemical cocktail during the COVID-19 outbreak in Beijing, China: Insights from six-year aerosol particle composition measurements during the Chinese New Year holiday.

The rapidly spread coronavirus disease (COVID-19) has limited people's outdoor activities and hence caused substantial reductions in anthropogenic emissions around the world. However, the air quality in some megacities has not been improved as expected due to the complex responses of aerosol chemistry to the changes in precursors and meteorology. Here we demonstrate the responses of primary and secondary aerosol species to the changes in anthropogenic emissions during the COVID-19 outbreak in Beijing, China along with the Chinese New Year (CNY) holiday effects on air pollution by using six-year aerosol particle composition measurements. Our results showed large reductions in primary aerosol species associated with traffic, cooking and coal combustion emissions by 30-50% on average during the CNY, while the decreases in secondary aerosol species were much small (5-12%). These results point towards a future challenge in mitigating secondary air pollution because the reduced gaseous precursors may not suppress secondary aerosol formation efficiently under stagnant meteorological conditions. By analyzing the long-term measurements from 2012 to 2020, we found considerable increases in the ratios of nitrate to sulfate, secondary to primary OA, and sulfur and nitrogen oxidation capacity despite the overall decreasing trends in mass concentrations of most aerosol species, suggesting that the decreases in anthropogenic emissions have facilitated secondary formation processes during the last decade. Therefore, a better understanding of the mechanisms driving the chemical responses of secondary aerosol to the changes in anthropogenic emissions under complex meteorological environment is essential for future mitigation of air pollution in China.