This study was to optimize the current methods for identifying and predicting the risk of critical illness in patients with connective tissue disease-associated interstitial lung disease (CTD-ILD). First, 200 patients diagnosed with CTD-ILD were included, and detailed demographic, serological, and imaging data were collected. Second, a risk identification and prediction framework was constructed based on multivariate logistic regression and machine learning algorithms (random forest (RF) and convolutional neural network (CNN)) to identify significant determinants of critical illness. Finally, the overall performance of each model was evaluated using K-fold cross-validation and external validation procedures. A feature ablation experiment was conducted based on the optimal random forest model to validate the independent contribution of each core predictor. The results showed that the logistic regression, random forest (RF), and CNN models were all successfully constructed and validated, among which the RF model demonstrated the best overall performance, with an accuracy of 85.7%, an area under the curve (AUC) of 0.88, a sensitivity of 83.5%, and a specificity of 88.2%. The ablation experiment confirmed that each feature had independent predictive value, with the most significant decline in model performance observed after the removal of IL‑6. Among them, the individual AUC value of interleukin-6 (IL-6) reached 0.981. Significant risk factors included patient age, C-reactive protein (CRP) level, presence of honeycomb lung on imaging, and the ratio of arterial oxygen partial pressure to inhaled oxygen concentration (PaO2/FiO2). The model in this study demonstrated satisfactory predictive ability and stability in both internal and external validation phases. The random forest model performed excellently in predicting the likelihood of critical illness in patients with CTD-ILD.

Purpose To establish molecular-structural mismatch indexes and explore their potential for the preoperative risk stratification of individuals with pediatric neuroblastoma. Materials and Methods In this prospective study, 102 pediatric individuals with suspected neuroblastoma were initially enrolled and underwent structural MRI and amide proton transfer (APT) MRI from April 2019 to November 2023. A four-pool Lorentzian fitting model was employed to minimize motion artifacts and extract the APT signal from confounding factors, generating APT maps and APT-weighted (APTw) images. Novel molecular-structural mismatch images were then synthesized using APT maps, APTw images, and conventional structural MRI. The mean tumor region of interest values on APT, APTw, and mismatch images were compared across risk groups using Wilcoxon rank sum tests; receiver operating characteristic analysis was used to assess the performance of these values in risk stratification, and the DeLong test was used to evaluate differences in the area under the curve (AUC). Results The final analysis included 71 participants (mean ± SD age at examination, 45.8 months ± 31.6; 52 male individuals), comprising 42 participants at high risk and 29 at non-high risk. The mean tumor region of interest values on APT, APTw, and mismatch images were greater in those at high risk than those not at high risk (P < .05). The mean APTw value had an AUC of 0.68 (95% CI: 0.54, 0.80), and the APT map achieved an AUC of 0.76 (95% CI: 0.64, 0.87). The molecular-structural mismatch images had numerically higher AUC values, at 0.84 (95% CI: 0.75, 0.94) for APTw-T2w mismatch and 0.83 (95% CI: 0.74, 0.93) for APT-T2w mismatch, respectively, but there was no evidence of a difference (P > .05). Conclusion Molecular-structural mismatch demonstrates potential as a noninvasive approach for differentiating neuroblastoma risk groups among children. Keywords: Pediatrics, MRI, Statistics, Molecular Imaging-Cancer, Abdomen/GI Supplemental material is available for this article. © RSNA, 2026.

Gingival landmarks, cutoff values, and related variables for determining gingival phenotypes: A scoping review.

Objective The aim of this study was to evaluate the role of baseline oxygenation parameters, including PaO2/FiO2 and SpO2, as well as the inflammatory marker IL-6, in predicting in-hospital mortality among patients treated in the intensive care unit with a diagnosis of acute respiratory distress syndrome (ARDS). Methods This retrospective cohort study included 97 patients diagnosed with ARDS who were treated in the intensive care unit and had available baseline PaO2/FiO2, SpO2, and IL-6 measurements. Patients were divided into two groups according to clinical outcome: survivors and non-survivors. Baseline oxygenation and inflammatory parameters were compared between the groups. Factors associated with in-hospital mortality were evaluated using logistic regression analysis. The prognostic performance of individual parameters and the combined model was assessed using receiver operating characteristic (ROC) analysis. Results Among the 97 patients included in the study, 58 survived and 39 died during hospitalization, resulting in an overall in-hospital mortality rate of 40.2%. Non-survivors had significantly lower PaO2/FiO2 and SpO2 values compared with survivors, while IL-6 levels were significantly higher. In ROC analysis, PaO2/FiO2 showed the highest prognostic performance among individual parameters, with an area under the curve (AUC) of 0.944. The AUC values were 0.911 for SpO2 and 0.868 for IL-6. The combined model including PaO2/FiO2, SpO2, and IL-6 demonstrated the highest discrimination ability, with an AUC of 0.962. Conclusion Low baseline PaO2/FiO2 and SpO2 levels and elevated IL-6 values are strongly associated with in-hospital mortality in patients with ARDS. The combined use of oxygenation parameters and IL-6 may provide a simple, accessible, and clinically useful approach for early mortality risk stratification in intensive care settings.

Coronary heart disease(CHD) is an important cause of cardiovascular diseases and, under adverse conditions such as arteriosclerosis, will raise the risk of death significantly. Due to the shortcomings of traditional adiposity indices when assessing aerobic capacity (AC), this paper intended to explore whether the novel cardiometabolic index (CMI) could predict lower AC in patients with coronary heart disease. A retrospective analysis was performed on 378 hospitalized patients with coronary heart disease who underwent treatment and completed cardiopulmonary exercise testing. Patients were grouped according to AC levels and CMI levels, and the baseline characteristics of both groups were assessed.Three multivariable binary logistic regression models were constructed to evaluate the associations of the CMI, body mass index (BMI), and waist-to-height ratio (WHtR) with reduced AC. Receiver operating characteristic (ROC) curve analysis was performed to assess and compare the predictive performance of the three models, including the area under the curve (AUC). Model fit was further evaluated using the - 2 log-likelihood, Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). Calibration of each model was assessed using the Hosmer-Lemeshow test, and DeLong's test was used to compare differences in AUC values between models. Three multivariable logistic regression models were constructed, each incorporating different core variables (BMI, WHtR, and CMI) along with other relevant covariates. Variance inflation factor (VIF) analysis confirmed the absence of significant multicollinearity among variables. Model performance evaluation demonstrated that Model 3, which included CMI, had the best fit, with the lowest - 2 log-likelihood, AIC, and BIC values, good calibration (Hosmer-Lemeshow test, P = 0.441), and the highest AUC (0.671, 95% CI: 0.605-0.736). However, DeLong's test revealed that the difference in AUC between Model 3 and Model 1, as well as between Model 3 and Model 2, did not reach statistical significance (P = 0.062 and P = 0.110, respectively). CMI demonstrates clinical potential in evaluating the decline of AC in patients with CHD.

Evaluating diagnostic classifiers across Z-score–stratified regions: a localized ROC-based approach

Preoperative prediction of lymphovascular and perineural invasion in locally advanced gastric cancer via CT habitat analysis and deep learning: A dual-center study.

To identify diuretic resistance (DR) in patients with congestive heart failure (CHF) during intensive care treatment by analyzing the diuretic effect (DE) and to investigate their clinical characteristics and prognosis. Data of 1,744 patients with CHF in the MIMIC-IV database were analyzed. The trajectory of DE was examined by group-based trajectory modeling and its relationship with the diuretic dose-response then evaluated. The area under the curve (AUC) was used to assess the characteristics of DE at different time points of DR. The association between DE, DR, and clinical outcomes was investigated using logistic and Cox regression with different covariate adjustment strategies. The final model identified four trajectories of DE, among which Class 1 patients were identified as having DR, defined as having a minimal diuretic response of 3.819 mL/mg (95% CI 3.223-4.414, p < 0.001) before peak dosing, and minimal changes in diuretic adjustments. The DE at different time points effectively distinguished DR, with AUC values of 0.966 (95% CI 0.959-0.973) and 0.979 (95% CI 0.973-0.985) and optimal cut-off values of 6.515 and 12.557 at 6 h and 24 h, respectively. The DR group had significantly higher rates of in-hospital mortality (20.08% vs. 8.92%, p < 0.001), all-cause re-admission (23.11% vs. 17.08%, p = 0.012), and one-year mortality (40.45% vs. 26.26%, p < 0.001) compared to those observed in the non-DR group. The method of constructing DE trajectory models offers an effective approach to identify DR and provides novel insights for analyzing its characteristics and prognostic implications.

After total knee arthroplasty (TKA), 10-20% of patients remain unsatisfied. Well-performing clinical prediction models can provide individualized risk estimates and stratification in terms of poor outcomes, resulting in unnecessary surgeries being avoided and patients being counseled preoperatively. We aimed to create a precise, well-performing prediction model for clinical application using different machine learning algorithms to predict those patients who will have residual pain, a low total Oxford Knee Score (OKS) and the patient group who do not achieve minimally clinical important difference (MCID) in OKS 1 year after TKA. We conducted a retrospective cohort study based on patients who had undergone primary TKA at our institution combined with 751 patient-related variables. The multivariable models used were based on the results of univariate analysis. We used the machine learning method Extreme Gradient Boosting (XGBoost). The discrimination capability of the models was measured with the area under the curve (AUC). 11,755 patients were included in this study. There were 850 (7.2%) patients who experienced persistent pain 1 year after TKA. The AUC was 0.67. For the secondary outcomes, the AUC values were similar. The most important variables in the model were lower preoperative OKS, younger age, valgus malalignment, lower preoperative pain OKS, use of mild opioid, neuropathic pain medicine and thyroxine, and higher body mass index. The prediction models achieved poor AUCs. It seems clear that the prediction of pain and functional outcome after TKA is difficult, even with a large patient cohort combined with 751 patient-related variables and sophisticated machine-learning algorithms.

A diagnostic plasma omics\u2010biomarker for Alzheimer's disease informed by microglial single\u2010cell transcriptomics: A pilot study

Lower Trapezius Transfer Maintains Meaningful Outcomes at 5 Years Based on Procedure-Specific Minimal Clinically Important Difference and Patient Acceptable Symptom State Achievement.

Human saliva is a heterogeneous bodily fluid with a complex composition, which contains antibodies, proteins, and viruses, making it applicable in clinical diagnosis. There are several advantages of the analysis of saliva samples over other biofluids, including a non-invasive and simple collection procedure for extraoral detection. Biomarker or pathogen detection in saliva can be performed with various methods: mass spectrometry, PCR, ELISA, electrochemical, and optical methods such as fluorescence, SPR, and SERS. The early detection of cancer and other disease biomarkers, as well as infectious agents, can be crucial for effective treatment and minimization of mortality from those diseases. The following paper reviews extraoral detection techniques to identify the most sensitive methods for diagnosing early and asymptomatic patients. The LODs collected and tabulated from 149 analytical papers, alongside the sensitivity, specificity, and sometimes the area under the curve (AUC) tabulated from 118 clinical studies, have all become parameters for the comparative quantitative analysis. Based on the limited but substantial number of analytical studies on the detection of cortisol in saliva (29), the electrochemical platforms demonstrated the highest sensitivity, with a geometric mean LOD of 11 pM. Within these methods, voltametric ones showed the best performance with 6 pM geometric mean LOD. Electrochemical techniques are then followed by immunoassay- and mass spectrometry-based platforms, with corresponding geometric average LOD values of 39.1 and 171 pM, respectively. However, clinical outcomes are at least as meaningful as LOD values. In terms of clinical analysis, ELISA and direct-SERS outperformed other methods, achieving balanced accuracy of approximately 87% and AUC values of 0.96 for direct SERS and 0.86 for ELISA. MS and PCR followed closely, with balanced accuracies around 84%. While the direct SERS is not yet widespread in clinical applications, its potential can be forged if the standardization issue is addressed.

Vendor-neutral deep learning reconstruction of dynamic contrast-enhanced prostate MRI: Image quality improvement and preservation of diagnostic performance.

Due to immunosuppression, mucosal barrier injury, and prolonged neutropenia resulting from both the disease and chemotherapy, along with the frequent use of broad-spectrum antibiotics and glucocorticoids, children with leukemia are at a high risk of invasive fungal disease (IFD). The present study aimed to develop an effective machine learning model to predict fungal infections in children with leukemia. A total of 247 pediatric patients diagnosed with leukemia and concurrent infections were evaluated. Five distinct ML classifiers-Random Forest, Logistic Regression, Support Vector Machine (SVM), Naïve Bayes, and K-Nearest Neighbors-were employed to construct predictive classification models. These models were trained using three distinct feature sets: (1) clinical features exclusively, (2) imaging features exclusively, and (3) an integrated feature set comprising both clinical and imaging data. The predictive model was validated prospectively in an independent cohort of 61 patients. Model performance was evaluated through cross-validation techniques to ensure robustness and generalizability. To validate the clinical applicability of the ML models, their diagnostic performance was systematically compared against that of three radiologists with varying experience levels: Reader A (3 years), Reader B (6 years), and Reader C (11 years). Among the five classifiers evaluated, models using both clinical and imaging features consistently outperformed those relying solely on either clinical or imaging features. Notably, the SVM algorithm exhibited the highest overall predictive performance. Within the SVM algorithm, the validation set achieved the mean area under the curve (AUC) values of 0.825 with clinical features alone, 0.852 with imaging features alone, and 0.947 when both clinical and imaging features were combined. The corresponding mean AUC values for the test set were 0.777, 0.797, and 0.879. Furthermore, a comparative analysis between the classification results of the SVM model and the diagnostic assessments provided by three radiologists demonstrated that the SVM consistently outperformed the radiologists across key performance metrics. The SVM algorithm demonstrates robust efficacy in predicting fungal infections among pediatric patients diagnosed with leukemia. Within the predictive model, the variables that exhibited the greatest influence included pleural thickening, neutropenia, hormone therapy, CRP level, mediastinal lymphadenopathy, and the presence of pleural effusion.

Assessing the discriminatory ability of hemoglobin concentration to predict iron stores in Cambodian women: A systematic review and individual participant data meta-analysis.

Loneliness and onset of depressive symptoms in older adults: Evaluating the utility of the 1-item, 3-item, and 20-item versions of the university of California, Los Angeles loneliness scale.

This study aims to characterize age inflection points for hypertension, elevated fasting blood glucose, and dyslipidemia in a Chinese physical examination population, analyze gender differences and the impact of comorbid conditions on these inflection points, and inform the development of gender-specific precision screening strategies. Data from 263,824 health examination records collected at the General Hospital of the Northern Theater Command between January 2020 and May 2025 were included. The maximum Youden index was used to determine age inflection points for single diseases and comorbid conditions. Chi-square tests were employed to compare disease prevalence before and after the inflection points, and odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. Significant gender differences were observed in inflection points for single diseases. The overall inflection point for hypertension was 40.5 years (38.5 years in males, 43.5 years in females). After the inflection point, prevalence increased from 7.7% to 23.1% in males (OR = 3.59, 95%CI: 3.46-3.72) and from 2.9% to 11.2% in females (OR = 4.24, 95%CI: 4.00-4.50). The overall inflection point for elevated fasting blood glucose was 44.5 years (43.5 years in males, 50.5 years in females). After the inflection point, prevalence increased from 4.6% to 23.6% in males (OR = 6.39, 95%CI: 6.13-6.66) and from 3.1% to 14.5% in females (OR = 5.36, 95%CI: 5.08-5.66). The overall inflection point for dyslipidemia was 40.5 years (31.5 years in males, 45.5 years in females). After the inflection point, prevalence increased from 36.7% to 46.6% in males (OR = 1.50, 95%CI: 1.45-1.56) and from 14.4% to 36.9% in females (OR = 3.48, 95%CI: 3.37-3.60). All differences were statistically significant (P < 0.001). Compared to single-disease models, comorbid conditions showed an average increase of 0.12 in area under the curve (AUC). The combination of elevated fasting blood glucose and hypertension in females had the highest AUC (0.75, 95%CI: 0.74-0.76), with sensitivity of 69.1%, specificity of 69.5%, and Youden index of 0.39. The combination of dyslipidemia and elevated fasting blood glucose achieved the highest Youden index (0.41), with significantly higher AUC in females (0.76, 95%CI: 0.75-0.77) than in males (0.72, 95%CI: 0.71-0.72). For triple comorbidity, AUC was 0.75 (95%CI: 0.73-0.76) in females versus 0.70 (95%CI: 0.69-0.71) in males. Female inflection points for comorbidities were consistently later than those in males, and AUC values were uniformly higher in females, with the most pronounced gender difference observed for the combination of dyslipidemia and hypertension (AUC = 0.73 in females vs. 0.64 in males). This study identified gender-specific age inflection points for hypertension, elevated fasting blood glucose, and dyslipidemia, and further characterized these thresholds for comorbid conditions. Comorbidity models showed improved predictive performance over single-disease models. A consistent gender pattern was observed: female inflection points occurred later than those in males but converged around age 50 for all comorbid conditions, while the discriminative power of age for male dyslipidemia was limited. These findings may inform the development of gender-specific, comorbidity-focused screening strategies.

Participation is a key rehabilitation outcome. However, there is limited evidence on the measurement properties of patient-reported outcome measures (PROMs) that assess participation in rehabilitation settings. Therefore, this study evaluated the test-retest reliability and responsiveness of two widely used PROMs: the Utrecht Scale for Evaluation of Rehabilitation - Participation (USER-P) Restriction subscale and the Patient-Reported Outcomes Measurement Information System Ability to Participate in Social Roles and Activities 4-item short form (PROMIS-APS-SF) in inpatient and outpatient settings. In this multicentre prospective cohort study, inpatients and outpatients completed PROMs at the start of rehabilitation (T0), after six months (T1), and two weeks thereafter (T2). Test-retest reliability (T1-T2) was evaluated using intraclass correlation coefficients (ICCs), Bland-Altman plots, and the smallest detectable change (SDC). Responsiveness (T0-T1) was examined using effect sizes, area under the curve (AUC), and the minimal important change (MIC) based on the Global Rating of Change scale. A total of 553 patients completed PROMs at T0-T1, of whom 168 also completed them at T2. Scores on both PROMs demonstrated sufficient test-retest reliability (ICC > 0.70) across both rehabilitation settings. Moderate to large effect sizes were found, except for the PROMIS-APS-SF scores in inpatients, which showed a small effect size. The USER-P Restriction scores achieved sufficient AUC values for inpatients (0.71) and outpatients (0.72). At group level, MIC values exceeded the SDC for both PROMs, but only few did at individual level. Scores on both PROMs appeared appropriate for evaluating participation outcomes at group level within rehabilitation settings, with the USER-P Restriction scores showing better responsiveness among inpatients. However, the use of these scores for evaluating individual participation goals seems limited.

This study aims to develop an optimal model for distinguishing seminoma from non-seminoma testicular tumors using machine learning classifiers based on multiparametric MRI radiomics. This multi-institutional study enrolled a total of 188 patients, including 83 with seminoma and 105 with non-seminoma. The cohort from Institution 1 (n = 137) served as the training and validation set, whereas the independent cohort from Institution 2 (n = 51) was designated as the test set. Manual segmentation of tumor regions of interest (ROIs) was performed by experienced researchers on DWI, ADC, T2WI, and CE-T1WI sequences. A comprehensive radiomics workflow was implemented, encompassing data standardization, dimensionality reduction, feature selection, and classification using six distinct machine learning classifiers. Predictive models were developed by integrating radiomics features extracted from individual sequences and multiple sequence combinations with machine learning algorithms. In parallel, clinical models were established through univariate and multivariate logistic regression to identify significant predictive factors. A combined model incorporating both radiomics signatures and clinical characteristics was subsequently developed. The discriminatory performance of all models was evaluated by comparing area under the curve (AUC) values using DeLong's test. Model 19 (DWI+T2WI+ADC+CE-T1W+Clinical/SVM) achieved the highest AUC values, reaching 0.934, 0.922, and 0.846 on the training, validation, and external test sets respectively. In terms of clinical-radiological characteristics, alpha-fetoprotein (AFP) and cystic necrosis are significant predictors. A combined model utilising DWI, ADC, T2WI, and contrast-enhanced T1WI imaging, alongside AFP and cystic necrosis, demonstrates high precision, adaptability, and robustness in distinguishing seminoma from non-seminoma testicular tumors.

To evaluate the diagnostic value of Node Reporting and Data System (Node-RADS) and apparent diffusion coefficient (ADC) values for identifying axillary lymph node metastasis (ALNM) in breast cancer, and to construct and validate a predictive model for ALNM evaluation. The Node-RADS scores for axillary lymph nodes (ALN) were retrospectively assessed. The ADC values of the corresponding lymph nodes (LN) and the primary tumors were measured to calculate the calibrated ADC (cADC) and relative ADC (rADC) values. A predictive model was developed based on the factors associated with ALNM that were identified in the univariate and multivariate analyses. The model was subsequently validated on an internal and external validation dataset. The diagnostic performance was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC). Cohen's Kappa analysis was used to evaluate inter-reader agreement. Seven hundred eighty-seven female breast cancer patients from Center 1 (mean age, 52.01 years ± 9.42) and 63 from Center 2 (mean age, 53.21 years ± 11.32) were included. Node-RADS exhibited good diagnostic performance in distinguishing ALNM, with a score greater than 2 being the optimal cutoff value. The model incorporating Node-RADS and cADC showed excellent predictive ability, achieving AUC values of 0.807 (95% CI: 0.751, 0.856) and 0.801 (95% CI: 0.681, 0.891) in the internal and external validation sets, respectively. Node-RADS provides a reliable method for the standardized assessment of ALNM. The combination of Node-RADS with ADC improves the diagnostic performance for ALNM in breast cancer. Question Axillary lymph node status significantly influences treatment strategies for breast cancer, yet there remains a lack of consensus regarding their radiological evaluation. Findings Both Node-RADS and ADC values are effective for distinguishing lymph node metastasis, and their combination improves the diagnostic performance for ALNM. Clinical relevance The predictive model that integrates Node-RADS and cADC serves as a simple and practical tool to assist clinicians in formulating personalized treatment strategies.