ABSTRACT
PURPOSE: Identifying cancer symptoms in electronic health record (EHR) narratives is feasible with natural language processing (NLP). However, more efficient NLP systems are needed to detect various symptoms and distinguish observed symptoms from negated symptoms and medication-related side effects. We evaluated the accuracy of NLP in (1) detecting 14 symptom groups (ie, pain, fatigue, swelling, depressed mood, anxiety, nausea/vomiting, pruritus, headache, shortness of breath, constipation, numbness/tingling, decreased appetite, impaired memory, disturbed sleep) and (2) distinguishing observed symptoms in EHR narratives among patients with cancer. METHODS: We extracted 902,508 notes for 11,784 unique patients diagnosed with cancer and developed a gold standard corpus of 1,112 notes labeled for presence or absence of 14 symptom groups. We trained an embeddings-augmented NLP system integrating human and machine intelligence and conventional machine learning algorithms. NLP metrics were calculated on a gold standard corpus subset for testing. RESULTS: The interannotator agreement for labeling the gold standard corpus was excellent at 92%. The embeddings-augmented NLP model achieved the best performance (F1 score = 0.877). The highest NLP accuracy was observed in pruritus (F1 score = 0.937) while the lowest accuracy was in swelling (F1 score = 0.787). After classifying the entire data set with embeddings-augmented NLP, we found that 41% of the notes included symptom documentation. Pain was the most documented symptom (29% of all notes) while impaired memory was the least documented (0.7% of all notes). CONCLUSION: We illustrated the feasibility of detecting 14 symptom groups in EHR narratives and showed that an embeddings-augmented NLP system outperforms conventional machine learning algorithms in detecting symptom information and differentiating observed symptoms from negated symptoms and medication-related side effects.
Subject(s)
Electronic Health Records , Natural Language Processing , Neoplasms , Humans , Neoplasms/diagnosis , Neoplasms/psychology , Female , Male , Algorithms , Machine Learning , Narration , Middle AgedABSTRACT
INTRODUCTION: Malignant ventricular arrhythmia (VA) and sudden cardiac death (SCD) have been reported in patients with mitral valve prolapse (MVP); however, effective risk stratification methods are still lacking. Myocardial fibrosis is thought to play an important role in the development of VA; however, observational studies have produced contradictory findings regarding the relationship between VA and late gadolinium enhancement (LGE) in MVP patients. The aim of this meta-analysis and systematic review of observational studies was to investigate the association between left ventricular LGE and VA in patients with MVP. METHODS: We searched the PubMed, Embase, and Web of Science databases from 1993 to 2023 to identify case-control, cross-sectional, and cohort studies that compared the incidence of VA in patients with MVP who had left ventricular LGE and those without left ventricular LGE. RESULTS: A total of 1464 subjects with MVP from 12 observational studies met the eligibility criteria. Among them, VA episodes were reported in 221 individuals (15.1%). Meta-analysis demonstrated that the presence of left ventricular LGE was significantly associated with an increased risk of VA (pooled risk ratio 2.96, 95% CI: 2.26-3.88, p for heterogeneity = 0.07, I2 = 40%). However, a meta-regression analysis of the prevalence of mitral regurgitation (MR) showed that the severity of MR did not significantly affect the association between the occurrence of LGE and VA (p = 0.079). CONCLUSION: The detection of LGE could be helpful for stratifying the risk of VA in patients with MVP.
Subject(s)
Contrast Media , Gadolinium , Magnetic Resonance Imaging, Cine , Mitral Valve Prolapse , Humans , Mitral Valve Prolapse/complications , Mitral Valve Prolapse/diagnosis , Mitral Valve Prolapse/epidemiology , Mitral Valve Prolapse/physiopathology , Gadolinium/pharmacology , Magnetic Resonance Imaging, Cine/methods , Arrhythmias, Cardiac/diagnosis , Arrhythmias, Cardiac/etiology , Arrhythmias, Cardiac/epidemiology , Risk Factors , Risk Assessment/methodsABSTRACT
Metabolic dysfunction-associated liver disease (MASLD) and steatohepatitis (MASH) are becoming the most common causes of chronic liver disease in the United States and worldwide due to the obesity and diabetes epidemics. It is estimated that by 2030 close to 100 million people might be affected and patients with type 2 diabetes are especially at high risk. Twenty to 30% of patients with MASLD can progress to MASH, which is characterized by steatosis, necroinflammation, hepatocyte ballooning, and in advanced cases, fibrosis progressing to cirrhosis. Clinically, it is recognized that disease progression in diabetic patients is accelerated and the role of various genetic and epigenetic factors, as well as cell-matrix interactions in fibrosis and stromal remodeling, have recently been recognized. While there has been great progress in drug development and clinical trials for MASLD/MASH, the complexity of these pathways highlights the need to improve diagnosis/early detection and develop more successful antifibrotic therapies that not only prevent but reverse fibrosis.
ABSTRACT
CONTEXT: Extracting cancer symptom documentation allows clinicians to develop highly individualized symptom prediction algorithms to deliver symptom management care. Leveraging advanced language models to detect symptom data in clinical narratives can significantly enhance this process. OBJECTIVE: This study uses a pretrained large language model to detect and extract cancer symptoms in clinical notes. METHODS: We developed a pretrained language model to identify cancer symptoms in clinical notes based on a clinical corpus from the Enterprise Data Warehouse for Research at a healthcare system in the Midwestern United States. This study was conducted in 4 phases:1 pretraining a Bio-Clinical BERT model on one million unlabeled clinical documents,2 fine-tuning Symptom-BERT for detecting 13 cancer symptom groups within 1112 annotated clinical notes,3 generating 180 synthetic clinical notes using ChatGPT-4 for external validation, and4 comparing the internal and external performance of Symptom-BERT against a non-pretrained version and six other BERT implementations. RESULTS: The Symptom-BERT model effectively detected cancer symptoms in clinical notes. It achieved results with a micro-averaged F1-score of 0.933, an AUC of 0.929 internally, and 0.831 and 0.834 externally. Our analysis shows that physical symptoms, like Pruritus, are typically identified with higher performance than psychological symptoms, such as anxiety. CONCLUSION: This study underscores the transformative potential of specialized pretraining on domain-specific data in boosting the performance of language models for medical applications. The Symptom-BERT model's exceptional efficacy in detecting cancer symptoms heralds a groundbreaking stride in patient-centered AI technologies, offering a promising path to elevate symptom management and cultivate superior patient self-care outcomes.
Subject(s)
Electronic Health Records , Neoplasms , Humans , Neoplasms/diagnosis , Neoplasms/therapy , Natural Language Processing , AlgorithmsABSTRACT
Background and objective: Stress hyperglycemia is common in critically ill patients and is associated with poor prognosis. Whether this association exists in pulmonary hypertension (PH) patients is unknown. The present cohort study investigated the association of stress hyperglycemia with 90-day all-cause mortality in intensive care unit (ICU) patients with PH. Methods: Data of the study population were extracted from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. A new index, the ratio of admission glucose to HbA1c (GAR), was used to evaluate stress hyperglycemia. The study population was divided into groups according to GAR quartiles (Q1-Q4). The outcome of interest was all-cause mortality within 90 days, which was considered a short-term prognosis. Result: A total of 53,569 patients were screened. Ultimately, 414 PH patients were enrolled; 44.2% were male, and 23.2% were admitted to the cardiac ICU. As the GAR increased from Q2 to Q4, the groups had lower creatinine levels, longer ICU stays, and a higher proportion of renal disease. After adjusting for confounding factors such as demographics, vital signs, and comorbidities, an elevated GAR was associated with an increased risk of 90-day mortality. Conclusion: Stress hyperglycemia assessed by the GAR was associated with increased 90-day mortality in ICU patients with PH.
Subject(s)
Hyperglycemia , Hypertension, Pulmonary , Humans , Male , Female , Cohort Studies , Hypertension, Pulmonary/etiology , Critical Illness , Hyperglycemia/etiology , ComorbidityABSTRACT
BACKGROUND: People with cancer frequently experience severe and distressing symptoms associated with cancer and its treatments. Predicting symptoms in patients with cancer continues to be a significant challenge for both clinicians and researchers. The rapid evolution of machine learning (ML) highlights the need for a current systematic review to improve cancer symptom prediction. OBJECTIVE: This systematic review aims to synthesize the literature that has used ML algorithms to predict the development of cancer symptoms and to identify the predictors of these symptoms. This is essential for integrating new developments and identifying gaps in existing literature. METHODS: We conducted this systematic review in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist. We conducted a systematic search of CINAHL, Embase, and PubMed for English records published from 1984 to August 11, 2023, using the following search terms: cancer, neoplasm, specific symptoms, neural networks, machine learning, specific algorithm names, and deep learning. All records that met the eligibility criteria were individually reviewed by 2 coauthors, and key findings were extracted and synthesized. We focused on studies using ML algorithms to predict cancer symptoms, excluding nonhuman research, technical reports, reviews, book chapters, conference proceedings, and inaccessible full texts. RESULTS: A total of 42 studies were included, the majority of which were published after 2017. Most studies were conducted in North America (18/42, 43%) and Asia (16/42, 38%). The sample sizes in most studies (27/42, 64%) typically ranged from 100 to 1000 participants. The most prevalent category of algorithms was supervised ML, accounting for 39 (93%) of the 42 studies. Each of the methods-deep learning, ensemble classifiers, and unsupervised ML-constituted 3 (3%) of the 42 studies. The ML algorithms with the best performance were logistic regression (9/42, 17%), random forest (7/42, 13%), artificial neural networks (5/42, 9%), and decision trees (5/42, 9%). The most commonly included primary cancer sites were the head and neck (9/42, 22%) and breast (8/42, 19%), with 17 (41%) of the 42 studies not specifying the site. The most frequently studied symptoms were xerostomia (9/42, 14%), depression (8/42, 13%), pain (8/42, 13%), and fatigue (6/42, 10%). The significant predictors were age, gender, treatment type, treatment number, cancer site, cancer stage, chemotherapy, radiotherapy, chronic diseases, comorbidities, physical factors, and psychological factors. CONCLUSIONS: This review outlines the algorithms used for predicting symptoms in individuals with cancer. Given the diversity of symptoms people with cancer experience, analytic approaches that can handle complex and nonlinear relationships are critical. This knowledge can pave the way for crafting algorithms tailored to a specific symptom. In addition, to improve prediction precision, future research should compare cutting-edge ML strategies such as deep learning and ensemble methods with traditional statistical models.
ABSTRACT
Plasma total homocysteine (tHcy) and kidney function are both associated with mortality risk, but the degree to which kidney function modifies the impact of tHcy on mortality remains unknown. This prospective cohort study included a total of 14,225 hypertensive adults. Cox proportional hazard regression was used to analyze the separate and combined association of tHcy and estimated glomerular filtration rate (eGFR) with all-cause and cause-specific mortality. Mediation analysis was conducted to explore the mediating effect of eGFR. During a median follow-up of 4.0 years, 805 deaths were identified, including 397 deaths from cardiovascular disease (CVD). There were significant, positive relationships of tHcy with all-cause mortality (per 5 µmol/L; HR: 1.09; 95% CI: 1.07, 1.11), CVD mortality (HR: 1.11; 95% CI: 1.08, 1.13), and non-CVD mortality (HR: 1.07; 95% CI: 1.04, 1.10). The proportions of eGFR mediating these relationships were 39.1%, 35.7%, and 49.7%, respectively. There were additive interactions between tHcy and eGFR. Compared with those with low tHcy (<15 µmol/L) and high eGFR (≥90 mL·min-1·1.73 m-2), participants with high tHcy (≥20 µmol/L) and low eGFR (<60 mL·min-1·1.73 m-2) had the highest risk of all-cause mortality (HR: 4.89; 95% CI: 3.81, 6.28), CVD mortality (HR: 5.80; 95% CI: 4.01, 8.40), and non-CVD mortality (HR: 4.25; 95% CI: 3.02, 5.97). In conclusion, among Chinese hypertensive adults, high tHcy and impaired kidney function were independently and jointly associated with higher risks of all-cause and cause-specific mortality. Importantly, kidney function explained most (nearly 40%) of the increased risk of mortality conferred by high tHcy.
Subject(s)
Glomerular Filtration Rate , Homocysteine , Hypertension , Humans , Homocysteine/blood , Male , Female , Middle Aged , Hypertension/mortality , Hypertension/physiopathology , Hypertension/blood , Prospective Studies , Aged , Cardiovascular Diseases/mortality , Cardiovascular Diseases/blood , Adult , Kidney/physiopathology , Cause of Death , Mediation AnalysisABSTRACT
Type 2 diabetes mellitus is a major risk factor for hepatocellular carcinoma (HCC). Changes in extracellular matrix (ECM) mechanics contribute to cancer development1,2, and increased stiffness is known to promote HCC progression in cirrhotic conditions3,4. Type 2 diabetes mellitus is characterized by an accumulation of advanced glycation end-products (AGEs) in the ECM; however, how this affects HCC in non-cirrhotic conditions is unclear. Here we find that, in patients and animal models, AGEs promote changes in collagen architecture and enhance ECM viscoelasticity, with greater viscous dissipation and faster stress relaxation, but not changes in stiffness. High AGEs and viscoelasticity combined with oncogenic ß-catenin signalling promote HCC induction, whereas inhibiting AGE production, reconstituting the AGE clearance receptor AGER1 or breaking AGE-mediated collagen cross-links reduces viscoelasticity and HCC growth. Matrix analysis and computational modelling demonstrate that lower interconnectivity of AGE-bundled collagen matrix, marked by shorter fibre length and greater heterogeneity, enhances viscoelasticity. Mechanistically, animal studies and 3D cell cultures show that enhanced viscoelasticity promotes HCC cell proliferation and invasion through an integrin-ß1-tensin-1-YAP mechanotransductive pathway. These results reveal that AGE-mediated structural changes enhance ECM viscoelasticity, and that viscoelasticity can promote cancer progression in vivo, independent of stiffness.
Subject(s)
Carcinoma, Hepatocellular , Disease Progression , Elasticity , Extracellular Matrix , Liver Cirrhosis , Liver Neoplasms , Animals , Humans , beta Catenin/metabolism , Carcinoma, Hepatocellular/complications , Carcinoma, Hepatocellular/metabolism , Carcinoma, Hepatocellular/pathology , Cell Proliferation , Collagen/chemistry , Collagen/metabolism , Computer Simulation , Diabetes Mellitus, Type 2/complications , Diabetes Mellitus, Type 2/metabolism , Extracellular Matrix/metabolism , Glycation End Products, Advanced/metabolism , Integrin beta1/metabolism , Liver Neoplasms/complications , Liver Neoplasms/metabolism , Liver Neoplasms/pathology , Neoplasm Invasiveness , Viscosity , YAP-Signaling Proteins/metabolism , Liver Cirrhosis/complications , Liver Cirrhosis/metabolism , Liver Cirrhosis/pathologyABSTRACT
Hepatocyte ferroptosis promotes the pathogenesis and progression of liver fibrosis. Salvianolic acid B (Sal B) exerts antifibrotic effects. However, the pharmacological mechanism and target has not yet been fully elucidated. In this study, liver fibrosis was induced by CCl4 in wild-type mice and hepatocyte-specific extracellular matrix protein 1 (Ecm1)-deficient mice, which were separately treated with Sal B, ferrostatin-1, sorafenib or cilengitide. Erastin- or CCl4-induced hepatocyte ferroptosis models with or without Ecm1 gene knockdown were evaluated in vitro. Subsequently, the interaction between Ecm1 and xCT and the binding kinetics of Sal B and Ecm1 were determined. We found that Sal B significantly attenuated liver fibrosis in CCl4-induced mice. Ecm1 deletion in hepatocytes abolished the antifibrotic effect of Sal B. Mechanistically, Sal B protected against hepatocyte ferroptosis by upregulating Ecm1. Further research revealed that Ecm1 as a direct target for treating liver fibrosis with Sal B. Interestingly, Ecm1 interacted with xCT to regulate hepatocyte ferroptosis. Hepatocyte ferroptosis in vitro was significantly attenuated by Sal B treatment, which was abrogated after knockdown of Ecm1 in LO2 cells. Therefore, Sal B alleviates liver fibrosis in mice by targeting up-regulation of Ecm1 and inhibiting hepatocyte ferroptosis. The interaction between Ecm1 and xCT regulates hepatocyte ferroptosis.