Big data, we have all heard, promise to transform health care. But in the “hype cycle” of emerging technologies, machine learning now rides atop the “peak of inflated expectations,” and we need to ...better appreciate the technology’s capabilities and limitations.
Big data, we have all heard, promise to transform health care with the widespread capture of electronic health records and high-volume data streams from sources ranging from insurance claims and registries to personal genomics and biosensors.
1
Artificial-intelligence and machine-learning predictive algorithms, which can already automatically drive cars, recognize spoken language, and detect credit card fraud, are the keys to unlocking the data that can precisely inform real-time decisions. But in the “hype cycle” of emerging technologies, machine learning now rides atop the “peak of inflated expectations.”
2
Prediction is not new to medicine. From risk scores to guide anticoagulation (CHADS2) and . . .
Since 1999, more than 600 000 people have died from opioid overdoses in the USA and Canada, and the current rate of mortality in each country exceeds even that of the worst year of the HIV/AIDS ...epidemic. The Commission's analysis of the crisis focused on seven domains: (1) the North American opioid crisis as a case study in multi-system regulatory failure, (2) opioids' dual nature as both a benefit and a risk to health, (3) building integrated, well supported, and enduring systems of care for people with substance use disorders, (4) maximising the benefits and minimising the adverse effects of the involvement of the criminal justice system with people who are addicted to opioids, (5) creating healthy environments that can yield long-term declines in the incidence of addiction, (6) stimulating greater innovation in the response to the opioid crisis, and (7) preventing the North American opioid crisis from spreading globally. The Commission therefore recommends curtailing pharmaceutical product promotion, insulating medical education from pharmaceutical industry influence, closing the so-called revolving door between regulators and industry, making post-approval drug monitoring and risk mitigation a function of government, and firewalling bodies with formal power over prescribing from industry influence. ...a major investment in
To describe the infection control preparedness measures undertaken for coronavirus disease (COVID-19) due to SARS-CoV-2 (previously known as 2019 novel coronavirus) in the first 42 days after ...announcement of a cluster of pneumonia in China, on December 31, 2019 (day 1) in Hong Kong.
A bundled approach of active and enhanced laboratory surveillance, early airborne infection isolation, rapid molecular diagnostic testing, and contact tracing for healthcare workers (HCWs) with unprotected exposure in the hospitals was implemented. Epidemiological characteristics of confirmed cases, environmental samples, and air samples were collected and analyzed.
From day 1 to day 42, 42 of 1,275 patients (3.3%) fulfilling active (n = 29) and enhanced laboratory surveillance (n = 13) were confirmed to have the SARS-CoV-2 infection. The number of locally acquired case significantly increased from 1 of 13 confirmed cases (7.7%, day 22 to day 32) to 27 of 29 confirmed cases (93.1%, day 33 to day 42; P < .001). Among them, 28 patients (66.6%) came from 8 family clusters. Of 413 HCWs caring for these confirmed cases, 11 (2.7%) had unprotected exposure requiring quarantine for 14 days. None of these was infected, and nosocomial transmission of SARS-CoV-2 was not observed. Environmental surveillance was performed in the room of a patient with viral load of 3.3 × 106 copies/mL (pooled nasopharyngeal and throat swabs) and 5.9 × 106 copies/mL (saliva), respectively. SARS-CoV-2 was identified in 1 of 13 environmental samples (7.7%) but not in 8 air samples collected at a distance of 10 cm from the patient's chin with or without wearing a surgical mask.
Appropriate hospital infection control measures was able to prevent nosocomial transmission of SARS-CoV-2.
Protein kinase B (also known as AKT) and the mechanistic target of rapamycin (mTOR) are central regulators of T cell differentiation, proliferation, metabolism, and survival. Here, we show that ...during chronic murine lymphocytic choriomeningitis virus infection, activation of AKT and mTOR are impaired in antiviral cytotoxic T lymphocytes (CTLs), resulting in enhanced activity of the transcription factor FoxO1. Blockade of inhibitory receptor programmed cell death protein 1 (PD-1) in vivo increased mTOR activity in virus-specific CTLs, and its therapeutic effects were abrogated by the mTOR inhibitor rapamycin. FoxO1 functioned as a transcriptional activator of PD-1 that promoted the differentiation of terminally exhausted CTLs. Importantly, FoxO1-null CTLs failed to persist and control chronic viral infection. Collectively, this study shows that CTLs adapt to persistent infection through a positive feedback pathway (PD-1→FoxO1→PD-1) that functions to both desensitize virus-specific CTLs to antigen and support their survival during chronic viral infection.
Display omitted
•TCR activation of AKT and mTOR is impaired in exhausted CTLs•Therapeutic blockade of PD-1 signaling is mTOR dependent•FoxO1 is necessary to sustain CTL responses and control chronic viral infection•FoxO1 regulates the expression and differentiation of PD-1hi exhausted CTLs
Chronic infection can lead to a state of CD8+ T cell dysfunction referred to as exhaustion. Kaech and colleagues show that the transcription factor FoxO1 promotes the differentiation and maintenance of exhausted CD8+ T cells to control chronic viral infection.
Abstract
Objective
To conduct a systematic scoping review of explainable artificial intelligence (XAI) models that use real-world electronic health record data, categorize these techniques according ...to different biomedical applications, identify gaps of current studies, and suggest future research directions.
Materials and Methods
We searched MEDLINE, IEEE Xplore, and the Association for Computing Machinery (ACM) Digital Library to identify relevant papers published between January 1, 2009 and May 1, 2019. We summarized these studies based on the year of publication, prediction tasks, machine learning algorithm, dataset(s) used to build the models, the scope, category, and evaluation of the XAI methods. We further assessed the reproducibility of the studies in terms of the availability of data and code and discussed open issues and challenges.
Results
Forty-two articles were included in this review. We reported the research trend and most-studied diseases. We grouped XAI methods into 5 categories: knowledge distillation and rule extraction (N = 13), intrinsically interpretable models (N = 9), data dimensionality reduction (N = 8), attention mechanism (N = 7), and feature interaction and importance (N = 5).
Discussion
XAI evaluation is an open issue that requires a deeper focus in the case of medical applications. We also discuss the importance of reproducibility of research work in this field, as well as the challenges and opportunities of XAI from 2 medical professionals’ point of view.
Conclusion
Based on our review, we found that XAI evaluation in medicine has not been adequately and formally practiced. Reproducibility remains a critical concern. Ample opportunities exist to advance XAI research in medicine.
Clinical decision support systems that use artificial intelligence (AI) to improve diagnostic accuracy, efficiency, and safety have long been aspirational goals for computer scientists and ...clinicians. Yet diagnostic AI development has seen multiple cycles of inflated peaks of expectations followed by troughs of disillusionment. Clinicians are understandably wary of embracing new diagnostic AI solutions without understanding how they work and relate to their existing practice. Here, Chen et al describe learning methods that form the basis of many systems of diagnostic AI systems--learning from experts, examples, and experience--and their parallels to clinicians' existing approaches to learning.
Lineage tracing provides key insights into the fate of individual cells in complex organisms. Although effective genetic labeling approaches are available in model systems, in humans, most approaches ...require detection of nuclear somatic mutations, which have high error rates, limited scale, and do not capture cell state information. Here, we show that somatic mutations in mtDNA can be tracked by single-cell RNA or assay for transposase accessible chromatin (ATAC) sequencing. We leverage somatic mtDNA mutations as natural genetic barcodes and demonstrate their utility as highly accurate clonal markers to infer cellular relationships. We track native human cells both in vitro and in vivo and relate clonal dynamics to gene expression and chromatin accessibility. Our approach should allow clonal tracking at a 1,000-fold greater scale than with nuclear genome sequencing, with simultaneous information on cell state, opening the way to chart cellular dynamics in human health and disease.
Display omitted
•Somatic mtDNA mutations can track cellular relationships and hierarchies in vitro•Single-cell genomic assays faithfully detect mtDNA mutations•Lineage inference can be combined with gene expression or chromatin state profiles•mtDNA mutations enable studies of clonal architecture in human health and disease
Using single-cell sequencing technologies, somatic mutations in mtDNA can be used as natural genetic barcodes to study cellular states and clonal dynamics.
We propose an interpretable and scalable model to predict likely diagnoses at an encounter based on past diagnoses and lab results. This model is intended to aid physicians in their interaction with ...the electronic health records (EHR). To accomplish this, we retrospectively collected and de-identified EHR data of 2,701,522 patients at Stanford Healthcare over a time period from January 2008 to December 2016. A population-based sample of patients comprising 524,198 individuals (44% M, 56% F) with multiple encounters with at least one frequently occurring diagnosis codes were chosen. A calibrated model was developed to predict ICD-10 diagnosis codes at an encounter based on the past diagnoses and lab results, using a binary relevance based multi-label modeling strategy. Logistic regression and random forests were tested as the base classifier, and several time windows were tested for aggregating the past diagnoses and labs. This modeling approach was compared to a recurrent neural network based deep learning method. The best model used random forest as the base classifier and integrated demographic features, diagnosis codes, and lab results. The best model was calibrated and its performance was comparable or better than existing methods in terms of various metrics, including a median AUROC of 0.904 (IQR 0.838, 0.954) over 583 diseases. When predicting the first occurrence of a disease label for a patient, the median AUROC with the best model was 0.796 (IQR 0.737, 0.868). Our modeling approach performed comparably as the tested deep learning method, outperforming it in terms of AUROC (p < 0.001) but underperforming in terms of AUPRC (p < 0.001). Interpreting the model showed that the model uses meaningful features and highlights many interesting associations among diagnoses and lab results. We conclude that the multi-label model performs comparably with RNN based deep learning model while offering simplicity and potentially superior interpretability. While the model was trained and validated on data obtained from a single institution, its simplicity, interpretability and performance makes it a promising candidate for deployment.