Use of Natural Language Processing to Infer Sites of Metastatic Disease From Radiology Reports at Scale.

PURPOSE: To evaluate natural language processing (NLP) methods to infer metastatic sites from radiology reports. METHODS: A set of 4,522 computed tomography (CT) reports of 550 patients with 14 types of cancer was used to fine-tune four clinical large language models (LLMs) for multilabel classification of metastatic sites. We also developed an NLP information extraction (IE) system (on the basis of named entity recognition, assertion status detection, and relation extraction) for comparison. Model performances were measured by F1 scores on test and three external validation sets. The best model was used to facilitate analysis of metastatic frequencies in a cohort study of 6,555 patients with 53,838 CT reports. RESULTS: The RadBERT, BioBERT, GatorTron-base, and GatorTron-medium LLMs achieved F1 scores of 0.84, 0.87, 0.89, and 0.91, respectively, on the test set. The IE system performed best, achieving an F1 score of 0.93. F1 scores of the IE system by individual cancer type ranged from 0.89 to 0.96. The IE system attained F1 scores of 0.89, 0.83, and 0.81, respectively, on external validation sets including additional cancer types, positron emission tomography-CT ,and magnetic resonance imaging scans, respectively. In our cohort study, we found that for colorectal cancer, liver-only metastases were higher in de novo stage IV versus recurrent patients (29.7% v 12.2%; P < .001). Conversely, lung-only metastases were more frequent in recurrent versus de novo stage IV patients (17.2% v 7.3%; P < .001). CONCLUSION: We developed an IE system that accurately infers metastatic sites in multiple primary cancers from radiology reports. It has explainable methods and performs better than some clinical LLMs. The inferred metastatic phenotypes could enhance cancer research databases and clinical trial matching, and identify potential patients for oligometastatic interventions.

Up to 206 Million People Reached and Over 5.4 Million Trained in Cardiopulmonary Resuscitation Worldwide: The 2019 International Liaison Committee on Resuscitation World Restart a Heart Initiative.

Sudden out-of-hospital cardiac arrest is the third leading cause of death in industrialized nations. Many of these lives could be saved if bystander cardiopulmonary resuscitation rates were better. "All citizens of the world can save a life-CHECK-CALL-COMPRESS." With these words, the International Liaison Committee on Resuscitation launched the 2019 global "World Restart a Heart" initiative to increase public awareness and improve the rates of bystander cardiopulmonary resuscitation and overall survival for millions of victims of cardiac arrest globally. All participating organizations were asked to train and to report the numbers of people trained and reached. Overall, social media impact and awareness reached up to 206 million people, and >5.4 million people were trained in cardiopulmonary resuscitation worldwide in 2019. Tool kits and information packs were circulated to 194 countries worldwide. Our simple and unified global message, "CHECK-CALL-COMPRESS," will save hundreds of thousands of lives worldwide and will further enable many policy makers around the world to take immediate and sustainable action in this most important healthcare issue and initiative.

"Not just another walking program": Everyday Activity Supports You (EASY) model-a randomized pilot study for a parallel randomized controlled trial.

BACKGROUND: Maintaining physical activity is an important goal with positive health benefits, yet many people spend most of their day sitting. Our Everyday Activity Supports You (EASY) model aims to encourage movement through daily activities and utilitarian walking. The primary objective of this phase was to test study feasibility (recruitment and retention rates) for the EASY model. METHODS: This 6-month study took place in Vancouver, Canada, from May to December 2013, with data analyses in February 2014. Participants were healthy, inactive, community-dwelling women aged 55-70 years. We recruited through advertisements in local community newspapers and randomized participants using a remote web service. The model included the following: group-based education and social support, individualized physical activity prescription (called Activity 4-1-1), and use of a Fitbit activity monitor. The control group received health-related information only. The main outcome measures were descriptions of study feasibility (recruitment and retention rates). We also collected information on activity patterns (ActiGraph GT3X+ accelerometers) and health-related outcomes such as body composition (height and weight using standard techniques), blood pressure (automatic blood pressure monitor), and psychosocial variables (questionnaires). RESULTS: We advertised in local community newspapers to recruit participants. Over 3 weeks, 82 participants telephoned; following screening, 68% (56/82) met the inclusion criteria and 45% (25/56) were randomized by remote web-based allocation. This included 13 participants in the intervention group and 12 participants in the control group (education). At 6 months, 12/13 (92%) intervention and 8/12 (67%) control participants completed the final assessment. Controlling for baseline values, the intervention group had an average of 2,080 [95% confidence intervals (CIs) 704, 4,918] more steps/day at 6 months compared with the control group. There was an average between group difference in weight loss of -4.3 [95% CI -6.22, -2.40] kg and reduction in diastolic blood pressure of -8.54 [95% CI -16.89, -0.198] mmHg, in favor of EASY. CONCLUSIONS: The EASY pilot study was feasible to deliver; there was an increase in physical activity and reduction in weight and blood pressure for intervention participants at 6 months. TRIAL REGISTRATION: ClinicalTrials.gov identifier: NCT01842061.

Identification of residues in the receptor-binding domain (RBD) of the spike protein of human coronavirus NL63 that are critical for the RBD-ACE2 receptor interaction.

Human coronavirus NL63 (NL63), a member of the group I coronaviruses, may cause acute respiratory diseases in young children and immunocompromised adults. Like severe acute respiratory syndrome coronavirus (SARS-CoV), NL63 also employs the human angiotensin-converting enzyme 2 (hACE2) receptor for cellular entry. To identify residues in the spike protein of NL63 that are important for hACE2 binding, this study first generated a series of S1-truncated variants, examined their associations with the hACE2 receptor and subsequently mapped a minimal receptor-binding domain (RBD) that consisted of 141 residues (aa 476-616) towards the C terminus of the S1 domain. The data also demonstrated that the NL63 RBD bound to hACE2 more efficiently than its full-length counterpart and had a binding efficiency comparable to the S1 or RBD of SARS-CoV. A further series of RBD variants was generated using site-directed mutagenesis and random mutant library screening assays, and identified 15 residues (C497, Y498, V499, C500, K501, R518, R530, V531, G534, G537, D538, S540, E582, W585 and T591) that appeared to be critical for the RBD-hACE2 association. These critical residues clustered in three separate regions (designated RI, RII and RIII) inside the RBD, which may represent three receptor-binding sites. These results may help to delineate the molecular interactions between the S protein of NL63 and the hACE2 receptor, and may also enhance our understanding of the pathogenesis of NL63 and SARS-CoV.