External Validation Dataset Research Articles

Development and evaluation of multimodal AI for diagnosis and triage of ophthalmic diseases using ChatGPT and anterior segment images: protocol for a two-stage cross-sectional study.

Artificial intelligence (AI) technology has made rapid progress for disease diagnosis and triage. In the field of ophthalmic diseases, image-based diagnosis has achieved high accuracy but still encounters limitations due to the lack of medical history. The emergence of ChatGPT enables human-computer interaction, allowing for the development of a multimodal AI system that integrates interactive text and image information. To develop a multimodal AI system using ChatGPT and anterior segment images for diagnosing and triaging ophthalmic diseases. To assess the AI system's performance through a two-stage cross-sectional study, starting with silent evaluation and followed by early clinical evaluation in outpatient clinics. Our study will be conducted across three distinct centers in Shanghai, Nanjing, and Suqian. The development of the smartphone-based multimodal AI system will take place in Shanghai with the goal of achieving ≥90% sensitivity and ≥95% specificity for diagnosing and triaging ophthalmic diseases. The first stage of the cross-sectional study will explore the system's performance in Shanghai's outpatient clinics. Medical histories will be collected without patient interaction, and anterior segment images will be captured using slit lamp equipment. This stage aims for ≥85% sensitivity and ≥95% specificity with a sample size of 100 patients. The second stage will take place at three locations, with Shanghai serving as the internal validation dataset, and Nanjing and Suqian as the external validation dataset. Medical history will be collected through patient interviews, and anterior segment images will be captured via smartphone devices. An expert panel will establish reference standards and assess AI accuracy for diagnosis and triage throughout all stages. A one-vs.-rest strategy will be used for data analysis, and a post-hoc power calculation will be performed to evaluate the impact of disease types on AI performance. Our study may provide a user-friendly smartphone-based multimodal AI system for diagnosis and triage of ophthalmic diseases. This innovative system may support early detection of ocular abnormalities, facilitate establishment of a tiered healthcare system, and reduce the burdens on tertiary facilities. The study was registered in ClinicalTrials.gov on June 25th, 2023 (NCT05930444).

Dynamic rainfall-induced landslide susceptibility: A step towards a unified forecasting system

The initial inception of the landslide susceptibility concept defined it as a static property of the landscape, explaining the proneness of certain locations to generate slope failures. Since the spread of data-driven probabilistic solutions though, the original susceptibility definition has been challenged to incorporate dynamic elements that would lead the occurrence probability to change both in space and in time. This is the starting point of this work, which combines the traditional strengths of the susceptibility framework together with the strengths typical of landslide early warning systems. Specifically, we model landslide occurrences in the norther sector of Vietnam, using a multi-temporal landslide inventory recently released by NASA. A set of static (terrain) and dynamic (cumulated rainfall) covariates are selected to explain the landslide presence/absence distribution via a Bayesian version of a binomial Generalized Additive Models (GAM). Thanks to the large spatiotemporal domain under consideration, we include a large suite of cross-validation routines, testing the landslide prediction through random sampling, as well as through stratified spatial and temporal sampling. We even extend the model test towards regions far away from the study site, to be used as external validation datasets. The overall performance appears to be quite high, with Area Under the Curve (AUC) values in the range of excellent model results, and very few localized exceptions.This model structure may serve as the basis for a new generation of early warning systems. However, the use of The Climate Hazards group Infrared Precipitation with Stations (CHIRPS) for the rainfall component limits the model ability in terms of future prediction. Therefore, we envision subsequent development to take this direction and move towards a unified dynamic landslide forecast. Ultimately, as a proof-of-concept, we have also implemented a potential early warning system in Google Earth Engine.

Uncovering hub genes in sepsis through bioinformatics analysis.

In-depth studies on the mechanisms of pathogenesis of sepsis and diagnostic biomarkers in the early stages may be the key to developing individualized and effective treatment strategies. This study aimed to identify sepsis-related hub genes and evaluate their diagnostic reliability. The gene expression profiles of GSE4607 and GSE131761 were obtained from the Gene Expression Omnibus. Differentially co-expressed genes between the sepsis and control groups were screened. Single-sample gene set enrichment analysis and gene set variation analysis were performed to investigate the biological functions of the hub genes. A receiver operating characteristic curve was used to evaluate diagnostic value. Datasets GSE154918 and GSE185263 were used as external validation datasets to verify the reliability of the hub genes. Four differentially co-expressed genes, FAM89A, FFAR3, G0S2, and FGF13, were extracted using a weighted gene co-expression network analysis and differential gene expression analysis methods. These 4 genes were upregulated in the sepsis group and were distinct from those in the controls. Moreover, the receiver operating characteristic curves of the 4 genes exhibited considerable diagnostic value in discriminating septic blood samples from those of the non-septic control group. The reliability and consistency of these 4 genes were externally validated. Single-sample gene set enrichment analysis and gene set variation analysis analyses indicated that the 4 hub genes were significantly correlated with the regulation of immunity and metabolism in sepsis. The identified FAM89A, FFAR3, G0S2, and FGF13 genes may help elucidate the molecular mechanisms underlying sepsis and drive the introduction of new biomarkers to advance diagnosis and treatment.

Identification of angiogenesis-related genes in diabetic foot ulcer using machine learning algorithms

BackgroundDiabetic foot ulcers (DFUs) are among the most prevalent and dangerous complications of diabetes. Angiogenesis is pivotal for wound healing; however, its role in the chronic wound healing process in DFU requires further investigation. We aimed to investigate the pathogenic processes of angiogenesis in DFU from a molecular biology standpoint and to offer insightful information about DFU prevention and therapy. MethodsDifferential gene and weighted gene co-expression network analyses (WGCNA) were employed to screen for genes related to DFU using the downloaded and collated GSES147890 datasets. With the goal of identifying hub genes, an interaction among proteins (PPI) network was constructed, and enrichment analysis was carried out. Utilizing a variety of machine learning techniques, including Boruta, Support Vector Machine Recursive Feature Elimination (SVM-RFE), and Least Absolute Shrinkage and Selection Operator (LASSO), we were able to determine which hub genes most strongly correspond to DFU. This allowed us to create an ideally suited DFU forecasting model that was validated via an external dataset. Finally, by merging 36 angiogenesis-related genes (ARGs) and machine learning models, we identified the genes involved in DFU-related angiogenesis. ResultsBy merging 260 genes located in the green module and 59 differentially expressed genes (DEGs), 35 candidate genes highly associated with DFU were found for more investigation. 35 candidate genes were enriched in epidermal growth factor receptor binding, nuclear division regulation, fluid shear stress, atherosclerosis, and negative regulation of chromosomal structure for the enrichment study. Fifteen hub genes were found with the aid of the CytoHubba plug. The LASSO method scored better in terms of prediction performance (GSE134341) (LASSO:0.89, SVM:0.65, Boruta:0.66) based on the validation of the external datasets. We identified thrombomodulin (THBD) as a key target gene that potentially regulates angiogenesis during DFU development. Based on the external validation dataset (GSE80178 and GSE29221), receiver operating characteristic (ROC) curves with higher efficiency were generated to confirm the potential of THBD as a biomarker of angiogenesis in DFU. Furthermore supporting this finding were the results of Western blot and real-time quantitative polymerase chain reaction (RT-qPCR), which showed decreased THBD expression in human umbilical vein endothelial cells (HUVECs) cultivated under high glucose. ConclusionsThe findings implicate that THBD may influence DFU progression as a potential target for regulating angiogenesis, providing a valuable direction for future studies.

Development of predictive models for lymphedema by using blood tests and therapy data.

Lymphedema is a disease that refers to tissue swelling caused by an accumulation of protein-rich fluid that is usually drained through the lymphatic system. Detection of lymphedema is often based on expensive diagnoses such as bioimpedance spectroscopy, shear wave elastography, computed tomography, etc. In current machine learning models for lymphedema prediction, reliance on observable symptoms reported by patients introduces the possibility of errors in patient-input data. Moreover, these symptoms are often absent during the initial stages of lymphedema, creating challenges in its early detection. Identifying lymphedema before these observable symptoms manifest would greatly benefit patients by potentially minimizing the discomfort caused by these symptoms. In this study, we propose to use new data, such as complete blood count, serum, and therapy data, to develop predictive models for lymphedema. This approach aims to compensate for the limitations of using only observable symptoms data. We collected data from 2137 patients, including 356 patients with lymphedema and 1781 patients without lymphedema, with the lymphedema status of each patient confirmed by clinicians. The data for each patient included: (1) a complete blood count (CBC) test, (2) a serum test, and (3) therapy information. We used various machine learning algorithms (i.e. random forest, gradient boosting, decision tree, logistic regression, and artificial neural network) to develop predictive models on the training dataset (i.e. 80% of the data) and evaluated the models on the external validation dataset (i.e. 20% of the data). After selecting the best predictive models, we created a web application to aid medical doctors and clinicians in the rapid screening of lymphedema patients. A dataset of 2137 patients was assembled from Seoul National University Bundang Hospital. Predictive models based on the random forest algorithm exhibited satisfactory performance (balanced accuracy = 87.0 ± 0.7%, sensitivity = 84.3 ± 0.6%, specificity = 89.1 ± 1.5%, precision = 97.4 ± 0.7%, F1 score = 90.4 ± 0.4%, and AUC = 0.931 ± 0.007). We developed a web application to facilitate the swift screening of lymphedema among medical practitioners: https://snubhtxt.shinyapps.io/SNUBH_Lymphedema . Our study introduces a novel tool for the early detection of lymphedema and establishes the foundation for future investigations into predicting different stages of the condition.

Identification and immunological characterization of cuproptosis-related molecular clusters in ischemic stroke.

The present study elucidated cuproptosis-related molecular clusters involved in ischemic stroke and developed predictive models. Transcriptomic and immunological profiles of ischemic stroke-related datasets were extracted from the Gene Expression Omnibus database. Next, we conducted weighted gene co-expression network analysis to determine cluster-specific differentially expressed genes (DEGs). Models such as random forest and eXtreme gradient boosting (XGB) were evaluated to select the best prediction performance model. Subsequently, we validated the model's predictive efficiency by using nomograms, decision curve analysis, calibration curves, and receiver operating characteristic curve analysis with an external dataset. We identified two cuproptosis-related clusters involved in ischemic stroke. The DEGs in Cluster 2 were closely associated with amino acid metabolism, various immune responses, and cell proliferation pathways. The XGB model showed lower residuals, a smaller root mean square error, and a greater area under the curve value (AUC = 0.923), thus exhibiting the best discriminative performance. The AUC value for the external validation dataset was 0.921, thus confirming the high performance of the model. NFE2L2, NLRP3, GLS, LIPT1, and MTF1 were identified as potential cuproptosis predictors, thus shedding new light on ischemic stroke pathogenesis and heterogeneity.

Novel biomarkers to predict treatment response and prognosis in locally advanced rectal cancer undergoing neoadjuvant chemoradiotherapy

PurposeTo identify genes associated with treatment response and prognosis for locally advanced rectal cancer (LARC) patients receiving neoadjuvant chemoradiotherapy (NCRT).MethodsIn our cohort, gene expression profiles of 64 tumor biopsy samples before NCRT were examined and generated. Weighted gene co-expression network analysis was performed to identify gene modules. External validation datasets included GSE3493, GSE119409, and GSE133057. The expression of candidate genes was evaluated using immunohistochemistry (IHC). TIMER was used to assess immune infiltration.ResultsWe identified and validated the capability to predict the treatment response of CCT5 and ELF1 using our data and external validation datasets. The trends of survival differences of candidate genes in the GSE133057 dataset were similar to our cohort. High levels of CCT5 and ELF1 expression were associated with NCRT resistance and poor prognosis. Furthermore, the expression of CCT5 and ELF1 were also assessed in 117 LARC patients’ samples by the IHC method. Based on IHC results and Cox analysis, the risk score model with CCT5 and ELF1 was constructed and performed well. The risk score was an independent prognostic factor for progression-free survival and overall survival in LARC patients and was then used to build nomogram models. The underlying mechanisms of CCT5 and ELF1 were explored using gene set enrichment analysis. The underlying pathway including apoptosis, cell cycle, and other processes. CCT5 and ELF1 expressions were significantly correlated with immune cell infiltration.ConclusionCCT5 and ELF1 were determined as biomarkers for treatment response and prognosis in LARC patients. The risk score model and nomograms helped predict treatment response and survival outcomes for LARC patients undergoing NCRT.

Machine Learning Based Prediction of Postoperative Pancreatic Fistula Following Pancreaticoduodenectomy.

To develop a novel machine learning (ML) model to predict clinically relevant postoperative pancreatic fistula (CR-POPF) following pancreaticoduodenectomy (PD). Accurate prognostication of CR-POPF may allow for risk stratification and adaptive treatment strategies for potential PD candidates. However, antecedent models, such as the modified Fistula Risk Score (mFRS), are limited by poor discrimination and calibration. All records entailing PD within the 2014-2018 ACS NSQIP were identified. Additionally, patients undergoing PD at our institution between 2013 and 2021 were queried from our local data repository. An eXtreme Gradient Boosting (XGBoost) model was developed to estimate the risk of CR-POPF using data from the ACS NSQIP and evaluated using institutional data. Model discrimination was estimated using the area under the receiver operating characteristic (AUROC) and precision recall curve (AUPRC). Overall, 12,281 and 445 patients undergoing PD were identified within the 2014-2018 ACS NSQIP and our institutional registry, respectively. Application of the XGBoost and mFRS scores to the internal validation dataset revealed that the former model had significantly greater AUROC (0.72 vs. 0.68, P<0.001) and AUPRC (0.22 vs. 0.18, P<0.001). Within the external validation dataset, the XGBoost model remained superior to the mFRS with an AUROC of 0.79 (95% CI 0.74-0.84) versus 0.75 (95% CI 0.70-0.80, P<0.001). In addition, AUPRC was higher for the XGBoost model, compared to the mFRS. Our novel ML model consistently outperformed the previously validated mFRS within internal and external validation cohorts, thereby demonstrating its generalizability and utility for enhancing prediction of CR-POPF.

Can the generalizability issue of artificial intelligence be overcome? Pneumothorax detection algorithm.

The generalizability of artificial intelligence (AI) models is a major issue in the field of AI applications. Therefore, we aimed to overcome the generalizability problem of an AI model developed for a particular center for pneumothorax detection using a small dataset for external validation. Chest radiographs of patients diagnosed with pneumothorax (n = 648) and those without pneumothorax (n = 650) who visited the Ankara University Faculty of Medicine (AUFM; center 1) were obtained. A deep learning-based pneumothorax detection algorithm (PDA-Alpha) was developed using the AUFM dataset. For implementation at the Health Sciences University (HSU; center 2), PDA-Beta was developed through external validation of PDA-Alpha using 50 radiographs with pneumothorax obtained from HSU. Both PDA algorithms were assessed using the HSU test dataset (n = 200) containing 50 pneumothorax and 150 non-pneumothorax radiographs. We compared the results generated by the algorithms with those of physicians to demonstrate the reliability of the results. The areas under the curve for PDA-Alpha and PDA-Beta were 0.993 (95% confidence interval (CI): 0.985-1.000) and 0.986 (95% CI: 0.962-1.000), respectively. Both algorithms successfully detected the presence of pneumothorax on 49/50 radiographs; however, PDA-Alpha had seven false-positive predictions, whereas PDA-Beta had one. The positive predictive value increased from 0.525 to 0.886 after external validation (p = 0.041). The physicians' sensitivity and specificity for detecting pneumothorax were 0.585 and 0.988, respectively. The performance scores of the algorithms were increased with a small dataset; however, further studies are required to determine the optimal amount of external validation data to fully address the generalizability issue.

NIMG-21. SERIAL QUANTITATIVE MRI RADIOMICS SIGNATURE FOR PREDICTION OF SURVIVAL IN GLIOBLASTOMA – A PILOT STUDY

Abstract INTRODUCTION Radiomics in glioblastoma (GBM) imaging incorporates mathematical and statistical quantification of complex imaging features as surrogate biomarkers of tumour physiology. Delta-radiomics extends this capability by monitoring changes in these features over time. This study aims to evaluate selected radiomic and delta-radiomic features of diffusion-weighted imaging (DWI) as potential early predictors of overall survival (OS) in GBM patients undergoing adjuvant chemo-radiotherapy (aCRT). METHODS Imaging and data from n=12 newly diagnosed GBM patients prospectively recruited at Liverpool Hospital, New South Wales, were retrospectively analysed. Patients underwent serial MRI via 3.0T Siemens Magnetom-Skyra scanner, including anatomical and DWI sequences, before aCRT (baseline), and at 3 (FU1) and 6 weeks (FU2) post aCRT commencement. Gross tumour volume (GTV) was manually segmented on contrast-enhanced T1-weighted images. Four first-order histogram radiomic features (mean, skewness, entropy and 10th-percentile) and volume were extracted from the GTV on DWI sequences at each timepoint. Delta-radiomic features were calculated from the difference between measurements at baseline and each follow-up. Linear correlation between OS versus baseline and delta-radiomic features were calculated with Pearson’s coefficient (r). Statistically significant correlations were tested on an external validation dataset of n=8 GBM cases from The Cancer Imaging Archive (ACRIN-FMISO-Brain and ACRIN-DSC-MR-Brain datasets). RESULTS A significant positive correlation was found between OS and delta-entropy at FU2 (r = 0.639, p = 0.02). No significant correlation was found between OS and baseline or delta mean, skewness, or 10th percentile (p &amp;gt; 0.05). No significant correlation was found between OS and delta-volume at FU1 (r = 0.134, p = 0.68) or FU2 (r = 0.100, p = 0.76). CONCLUSION Our results indicate a potential utility of delta-entropy for prediction of overall survival. RESULTS will be validated on an external dataset. Further evaluation of radiomic and delta-radiomic features on a larger sample size is warranted.

Optimization and validation of 18F-DCFPyL PET radiomics-based machine learning models in intermediate- to high-risk primary prostate cancer.

Radiomics extracted from prostate-specific membrane antigen (PSMA)-PET modeled with machine learning (ML) may be used for prediction of disease risk. However, validation of previously proposed approaches is lacking. We aimed to optimize and validate ML models based on 18F-DCFPyL-PET radiomics for the prediction of lymph-node involvement (LNI), extracapsular extension (ECE), and postoperative Gleason score (GS) in primary prostate cancer (PCa) patients. Patients with intermediate- to high-risk PCa who underwent 18F-DCFPyL-PET/CT before radical prostatectomy with pelvic lymph-node dissection were evaluated. The training dataset included 72 patients, the internal validation dataset 24 patients, and the external validation dataset 27 patients. PSMA-avid intra-prostatic lesions were delineated semi-automatically on PET and 480 radiomics features were extracted. Conventional PET-metrics were derived for comparative analysis. Segmentation, preprocessing, and ML methods were optimized in repeated 5-fold cross-validation (CV) on the training dataset. The trained models were tested on the combined validation dataset. Combat harmonization was applied to external radiomics data. Model performance was assessed using the receiver-operating-characteristics curve (AUC). The CV-AUCs in the training dataset were 0.88, 0.79 and 0.84 for LNI, ECE, and GS, respectively. In the combined validation dataset, the ML models could significantly predict GS with an AUC of 0.78 (p<0.05). However, validation AUCs for LNI and ECE prediction were not significant (0.57 and 0.63, respectively). Conventional PET metrics-based models had comparable AUCs for LNI (0.59, p>0.05) and ECE (0.66, p>0.05), but a lower AUC for GS (0.73, p<0.05). In general, Combat harmonization improved external validation AUCs (-0.03 to +0.18). In internal and external validation, 18F-DCFPyL-PET radiomics-based ML models predicted high postoperative GS but not LNI or ECE in intermediate- to high-risk PCa. Therefore, the clinical benefit seems to be limited. These results underline the need for external and/or multicenter validation of PET radiomics-based ML model analyses to assess their generalizability.

Development and Validation of Data-Level Innovation Data-Balancing Machine Learning Models for Predicting Optimal Implantable Collamer Lens Size and Postoperative Vault.

There are only four sizes of implantable collamer lens (ICL) available for selection, which cannot completely fit all patients as a result of the discontinuity of ICL sizes. Sizing an optimal ICL and predicting postoperative vault are still unresolved problems. This study aimed to develop and validate innovative data-level data-balancing machine learning-based models for predicting ICL size and postoperative vault. The patients were randomly assigned to training and internal validation sets in a 4:1 ratio. Feature selection was performed using analysis of variance (ANOVA) and Kruskal-Wallis feature importance methods. Traditional linear regression model and machine learning-based models were used. The accuracy of models was assessed using the area under the curve (AUC) and confusion matrix. A total of 564 patients (1127 eyes) were eligible for this study, consisting of 808 eyes in the training set, 202 eyes in the internal validation set, and 117 eyes in the external validation set. Compared with the traditional linear regression method, the machine learning model bagging tree showed the best performance for ICL size selection, with an accuracy of 84.5% (95% confidence interval (CI) 83.2-85.8%), and the AUC ranged from 0.88 to 0.99; the prediction accuracy of 12.1mm and 13.7mm ICL sizes was improved by 49% and 59%, respectively. The bagging tree model achieved the best accuracy [90.2%, (95%CI 88.9-91.5%)] for predicting the postoperative vault, and the AUC ranged from 0.90 to 0.94. The prediction accuracies of internal and external validation dataset for ICL sizing were 82.2% (95%CI 81.1-83.3%) and 82.1% (95%CI 81.1-83.1%), respectively. The innovative data-level data balancing-based machine learning model can be used to predict ICL size and postoperative vault more accurately, which can assist surgeons in choosing optimal ICL size, thus reducing risks of postoperative complications and secondary surgery.

Digital image processing combined with machine learning: A new strategy for brown sugar classification

The coloring of foods is one of the main attributes of importance for consumers and it can be decisive for a consumer to accept or reject the product. Models that explore brown sugar coloring are scarce in scientific research. So, a new strategy for brown sugar classification through the combination of digital image processing, machine learning and physicochemical composition data was proposed. RGB channel intensities and color histogram data, obtained from digital image processing, in combination with some physicochemical characteristics (sucrose, Ca, Fe, ICUMSA color and total phenolic compounds (TPC)) were used as training and external validation datasets in the creation of classification models by RF algorithm. Excellent performance of classification models was observed by high overall accuracy rates for ICUMSA color (92.6 %), Ca and sucrose (100 %), Fe (94.9 %), and TPC (97.6 %). Thus, classifying brown sugar based on its color can be a valuable strategy for the beverage and food industries, allowing for greater diversification and meeting consumer needs while enhancing the quality and consistency of products.

Artificial intelligence-based classification of motor unit action potentials in real-world needle EMG recordings

ObjectiveTo develop an artificial neural network (ANN) for classification of motor unit action potential (MUAP) duration in real-word, unselected and uncleaned needle electromyography (n-EMG) recordings. MethodsTwo nested ANN models were trained, the first discerning muscle rest, contraction and artifacts in n-EMG recordings from 2674 individual muscles from 326 patients obtained as part of daily care. The second ANN model subsequently used segments labeled as contraction for prediction of prolonged, normal and shortened MUAPs. Model performance was assessed in one internal and two external validation datasets of 184, 30 and 50 muscles, respectively. ResultsThe first model discerned rest, contraction and artifacts with an accuracy of 96%. The second model predicted prolonged, normal and shortened MUAPs with an accuracy of 67%, 83% and 68% in the different validation sets. ConclusionsWe developed a two-step ANN that classifies rest, muscle contraction and artifacts from real-world n-EMG recordings with very high accuracy. MUAP duration classification had moderate accuracy. SignificanceThis is the first study to show that an ANN can classify MUAPs in real-world n-EMG recordings highlighting the potential for AI assisted MUAP classification as a clinical tool.

A New Risk Model based on the Machine Learning Approach for Prediction of Mortality in the Respiratory Intensive Care Unit.

Intensive care unit (ICU) resources are inadequate for the large population in China, so it is essential for physicians to evaluate the condition of patients at admission. In this study, our objective was to construct a machine-learning risk prediction model for mortality in respiratory intensive care units (RICUs). This study involved 817 patients who made 1,063 visits and who were admitted to the RICU from 2012 to 2017. Potential predictors such as demographic information, laboratory results, vital signs and clinical characteristics were considered. We constructed eXtreme Gradient Boosting (XGBoost) models and compared performances with random forest models, logistic regression models and clinical scores such as Acute Physiology and Chronic Health Evaluation II (APACHE II) and the sequential organ failure assessment (SOFA) system. The model was externally validated using data from Medical Information Mart for Intensive Care (MIMIC-III) database. A web-based calculator was developed for practical use. Among the 1,063 visits, the RICU mortality rate was 13.5%. The XGBoost model achieved the best performance with the area under the receiver operating characteristics curve (AUROC) of 0.860 (95% confidence interval (CI): 0.808 - 0.909) in the test set, which was significantly greater than APACHE II (0.749, 95% CI: 0.674 - 0.820; P = 0.015) and SOFA (0.751, 95% CI: 0.669 - 0.818; P = 0.018). The Hosmer-Lemeshow test indicated a good calibration of our predictive model in the test set with a P-value of 0.176. In the external validation dataset, the AUROC of XGBoost model was 0.779 (95% CI: 0.714 - 0.813). The final model contained variables that were previously known to be associated with mortality, but it also included some features absent from the clinical scores. The mean N-terminal pro-B-type natriuretic peptide (NTproBNP) of survivors was significantly lower than that of the non-survival group (2066.43 pg/mL vs. 8232.81 pg/mL; P < 0.001). Our results showed that the XGBoost model could be a suitable model for predicting RICU mortality with easy-to-collect variables at admission and help intensivists improve clinical decision-making for RICU patients. We found that higher NT-proBNP can be a good indicator of poor prognosis.

Development of a daily predictive model for the exacerbation of chronic obstructive pulmonary disease

Acute exacerbation (AE) of chronic obstructive pulmonary disease (COPD) compromises health status; it increases disease progression and the risk of future exacerbations. We aimed to develop a model to predict COPD exacerbation. We merged the Korean COPD subgroup study (KOCOSS) dataset with nationwide medical claims data, information regarding weather, air pollution, and epidemic respiratory virus data. The Korean National Health and Nutrition Examination Survey (KNHANES) dataset was used for validation. Several machine learning methods were employed to increase the predictive power. The development dataset consisted of 590 COPD patients enrolled in the KOCOSS cohort; these were randomly divided into training and internal validation subsets on the basis of the individual claims data. We selected demographic and spirometry data, medications for COPD and hospital visit for AE, air pollution data and meteorological data, and influenza virus data as contributing factors for the final model. Six machine learning and logistic regression tools were used to evaluate the performance of the model. A light gradient boosted machine (LGBM) afforded the best predictive power with an area under the curve (AUC) of 0.935 and an F1 score of 0.653. Similar favorable predictive performance was observed for the 2151 individuals in the external validation dataset. Daily prediction of the COPD exacerbation risk may help patients to rapidly assess their risk of exacerbation and will guide them to take appropriate intervention in advance. This might lead to reduction of the personal and socioeconomic burdens associated with exacerbation.

Identification and validation of novel prognostic signatures based on m5C methylation patterns and tumor EMT profiles in head and neck squamous cell carcinoma

The role of 5-methylcytosine (m5C) in tumor initiation and progression has been increasingly recognized. However, the precise association between the regulation of m5C and the progression, metastasis, and prognosis of head and neck squamous cell carcinoma (HNSCC) has not yet been fully explored. Data from 545 HNSCC patients obtained from The Cancer Genome Atlas (TCGA) database were analyzed. Unsupervised cluster analysis was conducted using the expression levels of m5C regulatory genes. Additionally, gene set variation analysis (GSVA), single-sample gene set enrichment analysis (ssGSEA), and Cox regression analysis were utilized. Quantitative reverse transcription polymerase chain reaction (RT-qPCR), colony formation assay, transwell experiments and western blots were performed in the HNSCC cell line UM-SCC-17B to assess the expression and functional role of one of the novel signatures, CNFN. Significant expression differences were found in m5C regulatory genes between tumor and normal tissues in HNSCC. Two distinct m5C modification patterns, characterized by substantial prognostic differences, were identified. Cluster-2, which exhibited a strong association with epithelial-mesenchymal transition (EMT), was found to be associated with a poorer prognosis. Based on the m5C clusters and EMT status, differentially expressed genes (DEGs) were identified. Using DEGs, an 8-gene signature (CAMK2N1, WNT7A, F2RL1, AREG, DEFB1, CNFN, TGFBI, and CAV1) was established to develop a prognostic model. The performance of this signature was validated in both the training and external validation datasets, demonstrating its promising efficacy. Furthermore, additional investigations using RT-qPCR on clinical specimens and experimental assays in cell lines provided compelling evidence suggesting that CNFN, one of the genes in the signature, could play a role in HNSCC progression and metastasis through the EMT pathway. This study highlighted the role of m5C in HNSCC progression and metastasis. The relationship between m5C and EMT has been elucidated for the first time. A robust prognostic model was developed for accurately predicting HNSCC patients’ survival outcomes. Potential molecular mechanisms underlying these associations have been illuminated through this research.

Elaboration and validation of a prognostic signature associated with disulfidoptosis in lung adenocarcinoma, consolidated with integration of single-cell RNA sequencing and bulk RNA sequencing techniques.

Lung adenocarcinoma (LUAD), the predominant subtype of non-small cell lung cancer (NSCLC), remains a pervasive global public health concern. Disulfidoptosis, a nascent form of regulated cell death (RCD), presents an emerging field of inquiry. Currently, investigations into disulfidoptosis are in their initial stages. Our undertaking sought to integrate single-cell RNA sequencing (scRNA-seq) in conjunction with traditional bulk RNA sequencing (bulk RNA-seq) methodologies, with the objective of delineating genes associated with disulfidoptosis and subsequently prognosticating the clinical outcomes of LUAD patients. Initially, we conducted an in-depth examination of the cellular composition disparities existing between LUAD and normal samples using scRNA-seq data sourced from GSE149655. Simultaneously, we scrutinized the expression patterns of disulfidoptosis-associated gene sets across diverse cell types. Subsequently, leveraging the bulk RNA-seq data, we formulated disulfidoptosis-related prognostic risk signatures (DRPS) employing LASSO-Cox regression. This was accomplished by focusing on genes implicated in disulfidoptosis that exhibited differential expression within endothelial cells (ECs). Sequentially, the robustness and precision of the DRPS model were rigorously verified through both internal and external validation datasets. In parallel, we executed single-cell trajectory analysis to delve into the differentiation dynamics of ECs. Concluding our study, we undertook a comprehensive investigation encompassing various facets. These included comparative assessments of enrichment pathways, clinicopathological parameters, immune cell abundance, immune response-associated genes, impacts of immunotherapy, and drug predictions among distinct risk cohorts. The scrutiny of scRNA-seq data underscored discernible disparities in cellular composition between LUAD and normal samples. Furthermore, disulfidoptosis-associated genes exhibited marked discrepancies within endothelial cells (ECs). Consequently, we formulated the Disulfidoptosis-Related Prognostic Signature (DRPS) to facilitate prognostic prediction. The prognostic nomogram based on the risk score effectively demonstrated DRPS's robust capacity to prognosticate survival outcomes. This assertion was corroborated by rigorous assessments utilizing both internal and external validation sets, thus affirming the commendable predictive accuracy and enduring stability of DRPS. Functional enrichment analysis shed light on the significant correlation of DRPS with pathways intrinsic to the cell cycle. Subsequent analysis unveiled correlations between DRPS and gene mutations characteristic of LUAD, as well as indications of an immunosuppressive status. Through drug prediction, we explored potential therapeutic agents for low-risk patients. Concluding our investigation, qRT-PCR experiments confirmed the heightened expression levels of EPHX1, LDHA, SHC1, MYO6, and TLE1 in lung cancer cell lines.

Development and validation of artificial intelligence models to predict urinary tract infections and secondary bloodstream infections in adult patients

BackgroundTraditional culture methods are time-consuming, making it difficult to utilize the results in the early stage of urinary tract infection (UTI) management, and automated urinalyses alone show insufficient performance for diagnosing UTIs. Several models have been proposed to predict urine culture positivity based on urinalysis. However, most of them have not been externally validated or consisted solely of urinalysis data obtained using one specific commercial analyzer. MethodsA total of 259,187 patients were enrolled to develop artificial intelligence (AI) models. AI models were developed and validated for the diagnosis of UTI and urinary tract related-bloodstream infection (UT-BSI). The predictive performance of conventional urinalysis and AI algorithms were assessed by the areas under the receiver operating characteristic curve (AUROC). We also visualized feature importance rankings as Shapley additive explanation bar plots. ResultsIn the two cohorts, the positive rates of urine culture tests were 25.2% and 30.4%, and the proportions of cases classified as UT-BSI were 1.8% and 1.6%. As a result of predicting UTI from the automated urinalysis, the AUROC were 0.745 (0.743–0.746) and 0.740 (0.737–0.743), and most AI algorithms presented excellent discriminant performance (AUROC > 0.9). In the external validation dataset, the XGBoost model achieved the best values in predicting both UTI (AUROC 0.967 [0.966–0.968]) and UT-BSI (AUROC 0.955 [0.951–0.959]). A reduced model using ten parameters was also derived. ConclusionsWe found that AI models can improve the early prediction of urine culture positivity and UT-BSI by combining automated urinalysis with other clinical information. Clinical utilization of the model can reduce the risk of delayed antimicrobial therapy in patients with nonspecific symptoms of UTI and classify patients with UT-BSI who require further treatment and close monitoring.

Development and Validation of Prediction Models for Exacerbation, Frequent Exacerbations and Severe Exacerbations of Chronic Obstructive Pulmonary Disease: A Registry Study in North China

In COPD patients, exacerbation has a detrimental influence on the quality of life, disease progression and socioeconomic burden. This study aimed to develop and validate models to predict exacerbation, frequent exacerbations and severe exacerbations in COPD patients. We conducted an observational prospective multicenter study. Clinical data of all outpatients with stable COPD were collected from Beijing Chaoyang Hospital and Beijing Renhe Hospital between January 2018 and December 2019. Patients were followed up for 1 year. The data from Chaoyang Hospital was used for modeling dataset, and that of Renhe Hospital was used for external validation dataset. The final dataset included 456 patients, with 326 patients as the model group and 130 patients as the validation group. Using LABA + ICS, frequent exacerbations in the past year and CAT score were independent risk factors for exacerbation in the next year (OR = 2.307, 2.722 and 1.147), and FVC %pred as a protective factor (OR = 0.975). Combined with chronic heart failure, frequent exacerbations in the past year, blood EOS counts and CAT score were independent risk factors for frequent exacerbations in the next year (OR = 4.818, 2.602, 1.015 and 1.342). Using LABA + ICS, combined with chronic heart failure, frequent exacerbations in the past year and CAT score were independent risk factors for severe exacerbations in the next year (OR = 1.950, 3.135, 2.980 and 1.133). Based on these prognostic models, nomograms were generated. The prediction models were simple and useful tools for predicting the risk of exacerbation, frequent exacerbations and severe exacerbations of COPD patients in North China.