An Interpretable Data-Driven Medical Knowledge Discovery Pipeline Based on Artificial Intelligence.

Difficulty in knowledge validation is a significant hindrance to knowledge discovery via data mining, especially automatic validation without artificial participation. In the field of medical research, medical knowledge discovery from electronic medical records is a common medical data mining method, but it is difficult to validate the discovered medical knowledge without the participation of medical experts. In this article, we propose a data-driven medical knowledge discovery closed-loop pipeline based on interpretable machine learning and deep learning; the components of the pipeline include Data Generator, Medical Knowledge Mining, Medical Knowledge Evaluation, and Medical Knowledge Application. In addition to completing the discovery of medical knowledge, the pipeline can also automatically validate the knowledge. We apply our pipeline's discovered medical knowledge to a traditional prognostic predictive model of heart failure in a real-world study, demonstrating that the incorporation of medical knowledge can effectively improve the performance of the traditional model. We also construct a scale model based on the discovered medical knowledge and demonstrate that it achieves good performance. To guarantee its medical effectiveness, every process of our pipeline involves the participation of medical experts.

The Role of "Hierarchical and Classified Prevention and Control Measures (HCPC)" Strategy for SARS-CoV-2 Delta Variant in Guangzhou: A Modeling Study.

BACKGROUND: The Delta variant of SARS-COV-2 has replaced previously circulating strains around the world in 2021. Sporadic outbreaks of the Delta variant in China have posed a concern about how to properly respond to the battle against evolving COVID-19. Here, we analyzed the "hierarchical and classified prevention and control (HCPC)" measures strategy deployed during the recent Guangzhou outbreak. METHODS: A modified susceptible-exposed-pre-symptomatic-infectious-recovered (SEPIR) model was developed and applied to study a range of different scenarios to evaluate the effectiveness of policy deployment. We simulated severe different scenarios to understand policy implementation and timing of implementation. Two outcomes were measured: magnitude of transmission and duration of transmission. The outcomes of scenario evaluations were presented relative to the reality case (i.e., 368 cases in 34 days) with 95% confidence interval (CI). RESULTS: Based on our simulation, the outbreak would become out of control with 7 million estimated infections under the assumption of the absence of any interventions than the 153 reported cases in reality in Guangzhou. The simulation on delayed implementation of interventions showed that the total case numbers would also increase by 166.67%-813.07% if the interventions were delayed by 3 days or 7 days. CONCLUSIONS: It may be concluded that timely and more precise interventions including mass testing and graded community management are effective measures for Delta variant containment in China.

A data-driven medical knowledge discovery framework to predict the length of ICU stay for patients undergoing craniotomy based on electronic medical records.

Craniotomy is an invasive operation with great trauma and many complications, and patients undergoing craniotomy should enter the ICU for monitoring and treatment. Based on electronic medical records (EMR), the discovery of high-risk multi-biomarkers rather than a single biomarker that may affect the length of ICU stay (LoICUS) can provide better decision-making or intervention suggestions for clinicians in ICU to reduce the high medical expenses of these patients and the medical burden as much as possible. The multi-biomarkers or medical decision rules can be discovered according to some interpretable predictive models, such as tree-based methods. Our study aimed to develop an interpretable framework based on real-world EMRs to predict the LoICUS and discover some high-risk medical rules of patients undergoing craniotomy. The EMR datasets of patients undergoing craniotomy in ICU were separated into preoperative and postoperative features. The paper proposes a framework called Rules-TabNet (RTN) based on the datasets. RTN is a rule-based classification model. High-risk medical rules can be discovered from RTN, and a risk analysis process is implemented to validate the rules discovered by RTN. The performance of the postoperative model was considerably better than that of the preoperative model. The postoperative RTN model had a better performance compared with the baseline model and achieved an accuracy of 0.76 and an AUC of 0.85 for the task. Twenty-four key decision rules that may have impact on the LoICUS of patients undergoing craniotomy are discovered and validated by our framework. The proposed postoperative RTN model in our framework can precisely predict whether the patients undergoing craniotomy are hospitalized for too long (more than 15 days) in the ICU. We also discovered and validated some key medical decision rules from our framework.

Application of big data and artificial intelligence in epidemic surveillance and containment.

Faced with the current time-sensitive COVID-19 pandemic, the overburdened healthcare systems have resulted in a strong demand to develop newer methods to control the spread of the pandemic. Big data and artificial intelligence (AI) have been leveraged amid the COVID-19 pandemic; however, little is known about their use for supporting public health efforts. In epidemic surveillance and containment, efforts are needed to treat critical patients, track and manage the health status of residents, isolate suspected cases, and develop vaccines and antiviral drugs. The applications of emerging practices of artificial intelligence and big data have become powerful "weapons" to fight against the pandemic and provide strong support in pandemic prevention and control, such as early warning, analysis and judgment, interruption and intervention of epidemic, to achieve goals of early detection, early report, early diagnosis, early isolation and early treatment. These are the decisive factors to control the spread of the epidemic and reduce the mortality. This paper systematically summarized the application of big data and AI in epidemic, and describes practical cases and challenges with emphasis on epidemic prevention and control. The included studies showed that big data and AI have the potential strength to fight against COVID-19. However, many of the proposed methods are not yet widely accepted. Thus, the most rewarding research would be on methods that promise value beyond COVID-19. More efforts are needed for developing standardized reporting protocols or guidelines for practice.

A model simulation on the SARS-CoV-2 Omicron variant containment in Beijing, China.

Objective: The Omicron variant of SARS-COV-2 is replacing previously circulating variants around the world in 2022. Sporadic outbreaks of the Omicron variant into China have posed a concern how to properly response to battle against evolving coronavirus disease 2019 (COVID-19). Methods: Based on the epidemic data from website announced by Beijing Center for Disease Control and Prevention for the recent outbreak in Beijing from April 22nd to June 8th in 2022, we developed a modified SEPIR model to mathematically simulate the customized dynamic COVID-zero strategy and project transmissions of the Omicron epidemic. To demonstrate the effectiveness of dynamic-changing policies deployment during this outbreak control, we modified the transmission rate into four parts according to policy-changing dates as April 22nd to May 2nd, May 3rd to 11st, May 12th to 21st, May 22nd to June 8th, and we adopted Markov chain Monte Carlo (MCMC) to estimate different transmission rate. Then we altered the timing and scaling of these measures used to understand the effectiveness of these policies on the Omicron variant. Results: The estimated effective reproduction number of four parts were 1.75 (95% CI 1.66-1.85), 0.89 (95% CI 0.79-0.99), 1.15 (95% CI 1.05-1.26) and 0.53 (95% CI 0.48 -0.60), respectively.  In the experiment, we found that till June 8th the cumulative cases would rise to 132,609 (95% CI 59,667-250,639), 73.39 times of observed cumulative cases number 1,807 if no policy were implemented on May 3rd, and would be 3,235 (95% CI 1,909 - 4,954), increased by 79.03% if no policy were implemented on May 22nd. A 3-day delay of the implementation of policies would led to increase of cumulative cases by 58.28% and a 7-day delay would led to increase of cumulative cases by 187.00%. On the other hand, taking control measures 3 or 7 days in advance would result in merely 38.63% or 68.62% reduction of real cumulative cases. And if lockdown implemented 3 days before May 3rd, the cumulative cases would be 289 (95% CI 211-378), reduced by 84%, and the cumulative cases would be 853 (95% CI 578-1,183), reduced by 52.79% if lockdown implemented 3 days after May 3rd. Conclusion: The dynamic COVID-zero strategy might be able to effectively minimize the scale of the transmission, shorten the epidemic period and reduce the total number of infections.

PaCAR: COVID-19 Pandemic Control Decision Making via Large-Scale Agent-Based Modeling and Deep Reinforcement Learning.

BACKGROUND: Policy makers are facing more complicated challenges to balance saving lives and economic development in the post-vaccination era during a pandemic. Epidemic simulation models and pandemic control methods are designed to tackle this problem. However, most of the existing approaches cannot be applied to real-world cases due to the lack of adaptability to new scenarios and micro representational ability (especially for system dynamics models), the huge computation demand, and the inefficient use of historical information. METHODS: We propose a novel Pandemic Control decision making framework via large-scale Agent-based modeling and deep Reinforcement learning (PaCAR) to search optimal control policies that can simultaneously minimize the spread of infection and the government restrictions. In the framework, we develop a new large-scale agent-based simulator with vaccine settings implemented to be calibrated and serve as a realistic environment for a city or a state. We also design a novel reinforcement learning architecture applicable to the pandemic control problem, with a reward carefully designed by the net monetary benefit framework and a sequence learning network to extract information from the sequential epidemiological observations, such as number of cases, vaccination, and so forth. RESULTS: Our approach outperforms the baselines designed by experts or adopted by real-world governments and is flexible in dealing with different variants, such as Alpha and Delta in COVID-19. PaCAR succeeds in controlling the pandemic with the lowest economic costs and relatively short epidemic duration and few cases. We further conduct extensive experiments to analyze the reasoning behind the resulting policy sequence and try to conclude this as an informative reference for policy makers in the post-vaccination era of COVID-19 and beyond. LIMITATIONS: The modeling of economic costs, which are directly estimated by the level of government restrictions, is rather simple. This article mainly focuses on several specific control methods and single-wave pandemic control. CONCLUSIONS: The proposed framework PaCAR can offer adaptive pandemic control recommendations on different variants and population sizes. Intelligent pandemic control empowered by artificial intelligence may help us make it through the current COVID-19 and other possible pandemics in the future with less cost both of lives and economy. HIGHLIGHTS: We introduce a new efficient, large-scale agent-based epidemic simulator in our framework PaCAR, which can be applied to train reinforcement learning networks in a real-world scenario with a population of more than 10,000,000.We develop a novel learning mechanism in PaCAR, which augments reinforcement learning with sequence learning, to learn the tradeoff policy decision of saving lives and economic development in the post-vaccination era.We demonstrate that the policy learned by PaCAR outperforms different benchmark policies under various reality conditions during COVID-19.We analyze the resulting policy given by PaCAR, and the lessons may shed light on better pandemic preparedness plans in the future.

A Phenogenetic Axis that Modulates Clinical Manifestation and Predicts Treatment Outcome in Primary Myeloid Neoplasms.

Although the concept of "myeloid neoplasm continuum" has long been proposed, few comparative genomics studies directly tested this hypothesis. Here we report a multi-modal data analysis of 730 consecutive newly diagnosed patients with primary myeloid neoplasm, along with 462 lymphoid neoplasm cases serving as the outgroup. Our study identified a "Pan-Myeloid Axis" along which patients, genes, and phenotypic features were all aligned in sequential order. Utilizing relational information of gene mutations along the Pan-Myeloid Axis improved prognostic accuracy for complete remission and overall survival in adult patients of de novo acute myeloid leukemia and for complete remission in adult patients of myelodysplastic syndromes with excess blasts. We submit that better understanding of the myeloid neoplasm continuum might shed light on how treatment should be tailored to individual diseases. Significance: The current criteria for disease diagnosis treat myeloid neoplasms as a group of distinct, separate diseases. This work provides genomics evidence for a "myeloid neoplasm continuum" and suggests that boundaries between myeloid neoplastic diseases are much more blurred than previously thought.

Coronavirus disease 2019 outbreak in Beijing's Xinfadi Market, China: a modeling study to inform future resurgence response.

BACKGROUND: A local coronavirus disease 2019 (COVID-19) case confirmed on June 11, 2020 triggered an outbreak in Beijing, China after 56 consecutive days without a newly confirmed case. Non-pharmaceutical interventions (NPIs) were used to contain the source in Xinfadi (XFD) market. To rapidly control the outbreak, both traditional and newly introduced NPIs including large-scale management of high-risk populations and expanded severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) PCR-based screening in the general population were conducted in Beijing. We aimed to assess the effectiveness of the response to the COVID-19 outbreak in Beijing's XFD market and inform future response efforts of resurgence across regions. METHODS: A modified susceptible-exposed-infectious-recovered (SEIR) model was developed and applied to evaluate a range of different scenarios from the public health perspective. Two outcomes were measured: magnitude of transmission (i.e., number of cases in the outbreak) and endpoint of transmission (i.e., date of containment). The outcomes of scenario evaluations were presented relative to the reality case (i.e., 368 cases in 34 days) with 95% Confidence Interval (CI). RESULTS: Our results indicated that a 3 to 14 day delay in the identification of XFD as the infection source and initiation of NPIs would have caused a 3 to 28-fold increase in total case number (31-77 day delay in containment). A failure to implement the quarantine scheme employed in the XFD outbreak for defined key population would have caused a fivefold greater number of cases (73 day delay in containment). Similarly, failure to implement the quarantine plan executed in the XFD outbreak for close contacts would have caused twofold greater transmission (44 day delay in containment). Finally, failure to implement expanded nucleic acid screening in the general population would have yielded 1.6-fold greater transmission and a 32 day delay to containment. CONCLUSIONS: This study informs new evidence that in form the selection of NPI to use as countermeasures in response to a COVID-19 outbreak and optimal timing of their implementation. The evidence provided by this study should inform responses to future outbreaks of COVID-19 and future infectious disease outbreak preparedness efforts in China and elsewhere.

PASCAL: a pseudo cascade learning framework for breast cancer treatment entity normalization in Chinese clinical text.

BACKGROUNDS: Knowledge discovery from breast cancer treatment records has promoted downstream clinical studies such as careflow mining and therapy analysis. However, the clinical treatment text from electronic health data might be recorded by different doctors under their hospital guidelines, making the final data rich in author- and domain-specific idiosyncrasies. Therefore, breast cancer treatment entity normalization becomes an essential task for the above downstream clinical studies. The latest studies have demonstrated the superiority of deep learning methods in named entity normalization tasks. Fundamentally, most existing approaches adopt pipeline implementations that treat it as an independent process after named entity recognition, which can propagate errors to later tasks. In addition, despite its importance in clinical and translational research, few studies directly deal with the normalization task in Chinese clinical text due to the complexity of composition forms. METHODS: To address these issues, we propose PASCAL, an end-to-end and accurate framework for breast cancer treatment entity normalization (TEN). PASCAL leverages a gated convolutional neural network to obtain a representation vector that can capture contextual features and long-term dependencies. Additionally, it treats treatment entity recognition (TER) as an auxiliary task that can provide meaningful information to the primary TEN task and as a particular regularization to further optimize the shared parameters. Finally, by concatenating the context-aware vector and probabilistic distribution vector from TEN, we utilize the conditional random field layer (CRF) to model the normalization sequence and predict the TEN sequential results. RESULTS: To evaluate the effectiveness of the proposed framework, we employ the three latest sequential models as baselines and build the model in single- and multitask on a real-world database. Experimental results show that our method achieves better accuracy and efficiency than state-of-the-art approaches. CONCLUSIONS: The effectiveness and efficiency of the presented pseudo cascade learning framework were validated for breast cancer treatment normalization in clinical text. We believe the predominant performance lies in its ability to extract valuable information from unstructured text data, which will significantly contribute to downstream tasks, such as treatment recommendations, breast cancer staging and careflow mining.