Prediction Score Research Articles

Risk factors for early mortality and impaired quality of life in oral cavity cancer - head and neck cancer register study.

Treatment of locoregionally advanced oral cavity cancer (OCC) is associated with treatment-related complications, functional deficits, and even early mortality. High-quality register data could help in choosing between curative and non-curative intent treatment options. The Helsinki Head and Neck Cancer Register (HHNCR) is linked with the EORTC QLQ-H&N35 questionnaire automatically sent to all patients at diagnosis and predetermined intervals. We analyzed pretreatment data of all patients diagnosed with OCC during 2018-2023, focusing on risk factors for early mortality and impaired health-related quality of life after curative-intent treatment. Of 597 patients, 556 (93%) were treated with curative intent. Thirty-nine (7.0%) patients died within 6 months after diagnosis. The independent risk-factors for 6-month mortality identified in multivariable analysis were T3 stage (OR 8.3 [2.6-26.5], p < 0.001), T4 stage (OR 8.2 [2.5-26.8], p < 0.001), N3 stage (OR 10.6 [3.2-35.1], p < 0.001), and Adult Comorbidity Evaluation (ACE)-27 score 2-3 (OR 5.5 [2.4-12.5], p < 0.001). These risk-factors were used to create a predictive risk score for early death. Younger, healthier patients had significantly higher EORTC QLQ-H&N35 response rates compared with older patients with comorbidities. Six months after diagnosis, patients with a stage III-IV tumor had significantly higher scores in 15 of 18 items, compared with patients with a stage I-II tumor. Early mortality was associated with advanced tumor (T) and nodal (N) stage, and increased pretreatment comorbidity (ACE-27) scores. The strongest predictor for impaired quality of life was locoregionally advanced disease.

Design and development of hybrid attention-based deep network for predicting water quality with weighted RBM features of heuristic approach

ABSTRACT Water resources are the main source to improve people's lives and enhance economic growth, which is connected to health and environmental practices. In addition to this, diverse deep learning models are employed for controlling water pollution. However, predicting the water quality effectively relies upon selecting the most appropriate features. So, the research work employs novel prediction techniques for accurately predicting the water quality. Here, the data is collected from publicly available sources and given into the feature extraction process using the restricted Boltzmann machine (RBM). The deep weighted features are evaluated optimally in RBM by the random parameter improved black widow optimization (RPIBO). Also, the hybrid attention-based deep network (HADNet) is designed to predict the water quality by considering the average prediction score. The experimental findings of dataset 1 and 2 of the suggested HADNet model achieve 95.7% and 96.8% regarding accuracy to control the water pollution.

Prediction Score for Major Bleeding in Patients With Venous Thromboembolism Receiving Direct Oral Anticoagulants - Insights From the COMMAND VTE Registry-2.

Predicting the bleeding risk during anticoagulation therapy is a key clinical challenge in patients with venous thromboembolism (VTE). However, there is no established prediction score for major bleeding (MB) in patients with VTE treated with direct oral anticoagulants (DOACs). Using the COMMAND VTE Registry-2, which enrolled 5,197 patients with acute symptomatic VTE between 2015 and 2020 among 31 centers in Japan, we investigated the risk factors for MB beyond 7 days and within 180 days in patients who received DOACs. A prediction score was developed in the derivation cohort (n=1,618), and prediction performance was evaluated in the validation cohort (n=809). Multivariate logistic regression analysis in the derivation cohort identified factors associated with MB. Based on β coefficients for each factor, the prediction score assigned 2 points to active cancer, history of MB, and thrombocytopenia, and 1 point to creatinine >1.2 mg/dL and anemia, summing them. The C statistic of the prediction score was 0.74 (95% confidence interval [CI] 0.68-0.80) in the derivation cohort and 0.74 (95% CI 0.67-0.81) in the validation cohort (P=0.98). When a cut-off value of 3 was used for the risk score, the sensitivity and specificity were 56.1% and 79.2%, respectively. The prediction score developed for MB during DOAC therapy (COMMAND-BLEED score) could be clinically useful for decision-making regarding anticoagulation strategies with DOACs.

Combining MRI radiomics, hypoxia gene signature score and clinical variables for prediction of biochemical recurrence-free survival after radiotherapy in prostate cancer.

To investigate the value of combining MRI radiomic and hypoxia-associated gene signature information with clinical data for predicting biochemical recurrence-free survival (BCRFS) after radiotherapy for prostate cancer. Patients with biopsy-proven prostate cancer, hypoxia-associated gene signature scores and pre-treatment MRI who received radiotherapy between 01/12/2007 and 31/08/2013 at two cancer centres were included in this retrospective cohort analysis. Prostate segmentation was performed on axial T2-weighted sequences using RayStation (v9.1). Histogram standardisation was applied prior to radiomic feature (RF) extraction. PyRadiomics (v3.0.1) was used to extract RFs for analysis. Four multivariable Cox proportional hazards BCRFS prediction models using clinical information alone and in combination with RFs and/or hypoxia scores were evaluated using concordance index (C-index) [confidence intervals (CI)]. Akaike Information Criterion (AIC) was used to assess model fit. 178 patients were included. The clinical-only model performance C-index score was 0.69 [0.64-0.7]. The combined clinical-radiomics model (C-index 0.70[0.66-0.73]) and clinical-radiomics-hypoxia model (C-index 0.70[0.65-0.73]) both had higher model performance. The clinical-hypoxia model (C-index 0.68 [0.63-0.7) had lower model performance. Based on AIC, addition of RFs to clinical variables alone improved model performance (p = 0.027), whereas adding hypoxia gene signature scores did not (p = 0.625). The selected features of the combined clinical-radiomics model included age, ISUP grade, tumour stage, and wavelet-derived grey level co-occurrence matrix (GLCM) RFs. Adding pre-treatment prostate MRI-derived radiomic features to a clinical model improves accuracy of predicting BCRFS after prostate radiotherapy, however addition of hypoxia gene signatures does not improve model accuracy.

Comparative Prognostic Utility of Updated Model for End-Stage Liver Disease Scores for Prediction of Early Mortality after Transjugular Intrahepatic Portosystemic Shunt Creation.

To compare the performance of updated model for end-stage liver disease (MELD) systems with that of the original MELD score for predicting early mortality after transjugular intrahepatic portosystemic shunt (TIPS) creation. In this single-center retrospective study, 6 MELD variations were quantified in 553 patients (n = 332; 60% male; mean age, 55 years) who underwent TIPS creation between 1998 and 2023. Scoring systems included original MELD, MELD-sodium (MELD-Na), MELD 3.0, MELD-lactate, MELD-glomerular filtration rate assessment in patients with liver disease-sodium (MELD-GRAIL-Na), and MELD-plus. Association of MELD schemes with 30-day, 6-week, and 90-day mortality was assessed using DeLong test, and the predictive capacity of MELD systems was evaluated by comparing area under receiver operating characteristic (AUROC) curves. TIPS were created for ascites (n = 263, 47%), variceal hemorrhage (n = 247, 45%), or other indications (n = 43, 8%). All MELD systems statistically associated with mortality at each time point (P < .001). Based on 30-day, 6-week, and 90-day AUROC curves, none of the updated MELD systems showed superior predictive capacity for early mortality compared with original MELD-MELD: 0.847, 0.841, and 0.818; MELD-Na : 0.847, 0.846, and 0.829; MELD 3.0: 0.848, 0.850, and 0.842; MELD-lactate: 0.915, 0.881, and 0.866; MELD-GRAIL-Na: 0.851, 0.847, and 0.831; and MELD-Plus: 0.843-0.898, 0.853-0.910, and 0.814-0.829, respectively (P > .05). Findings were principally confirmed on subset analyses stratified by sex, TIPS indication, TIPS urgency, stent type, and TIPS date. Updated MELD systems have prognostic value for early mortality after TIPS creation. However, in this study, these newer schemes did not offer additional predictive power beyond the original MELD, which still effectively estimates early post-TIPS survival.

Predicting hemorrhage expansion in patients with hypertensive intracerebral hemorrhage: the HE-VSD-A2TP score

BackgroundHematoma expansion (HE) in hypertensive intracerebral hemorrhage (HICH) is significantly associated with patient mortality. Early identification of HE would be planning for appropriate and aggressive management for improving outcome and containing HE. Existing HE prediction models show variable accuracy across settings. To address this limitation, we developed and validate a new predictive model to enhance the accuracy of HE in patients with HICH.MethodsWe conducted a retrospective cohort study using data from two centers. The primary outcome was the occurrence of HE within 24 h of symptom onset, defined as an increase in hematoma volume ≥33% or ≥12.5 mL on follow-up imaging. Logistic regression was used to identify independent predictors of HE, and the HE-VSD-A2TP score system was developed and validated.ResultsFive hundred and sixty seven patients in the derivation cohort and 378 patients in the validation cohort. The HE-VSD-A2TP score included age, uncontrolled blood pressure, hematoma volume, irregularity/lobulation shape, non-homogeneous density, presentation within 6 h from symptom onset to CT, and the use of anticoagulation/antiplatelet therapy. The HE-VSD-A2TP score demonstrated superior discrimination in predicting HE compared to existing models like PREDICT, 9-point, and BRAIN scores, with an AUC of 0.871(95%CI 0.839–0.904) in the derivation cohort and 0.858 (95%CI 0.819–0.897) in the validation cohort. The score also showed excellent calibration and outperformed other models in terms of sensitivity, specificity, likelihood ratios, negative predictive value, and positive predictive value. With regard to clinical usefulness, the decision curve analysis (DCA) of HE-VSD-A2TP showed higher net benefit than PREDICT, 9-point, and BRAIN scores in the both cohorts.ConclusionThe HE-VSD-A2TP score was validated to be an effective tool for identifying patients at risk of HE in patients with HICH. It was a valuable tool for guiding clinical management strategies and potentially improving patient outcomes.

Predicting Deterioration in Patients With Normotensive Acute Pulmonary Embolism Using Clinical-Imaging Features: AMulticenter Prospective Cohort Study.

Prioritization of management strategies in patients with normotensive acute pulmonary embolism is based on the identification of individuals at risk for early deterioration. This study aims to develop and validate a novel score for deterioration prediction using clinical-imaging features. This is multicenter, prospective observational cohort study (AOAPECT [Adverse Outcomes in Acute Pulmonary Embolism patients using Computed Tomography pulmonary angiography] cohort, NCT05098769). Registered-enrolled patients with normotensive acute pulmonary embolism were collected consecutively from 5 centers across China. Derivation set was established from 2 centers, while local and nonlocal external validation sets were derived from the remaining 3 centers. The end point was pulmonary embolism-related deterioration within 30 days after admission. Deterioration-related candidate predictors consisted of clinical, laboratory and computed tomography pulmonary angiography parameters were screened and then split into dichotomous values. The predictive score was conducted by a multivariable logistics regression and validated. Score performances were quantified using the area under the receiver operating characteristic curve. Of 3310 enrolled patients including 1 derivation (n=2061) and 2 validation sets (n=969 and 280), 272 patients (8.2%) experienced deterioration. In the derivation set, an increased risk of deterioration was observed with right to left ventricle diameter ratio ≥1.2, appearance of pulmonary vein sign on computed tomography pulmonary angiography, and heart rate ≥110 beats/min. When at least 2 out of 3 items were positive, patients were assigned to the high-risk deterioration group. This PE-RPHscore revealed good discrimination to deterioration in derivation and validation sets (area under the receiver operating characteristic curve, 0.82, 0.82, and 0.80). This PE-RPH score incorporating 2 computed tomography pulmonary angiography parameters and heart rate may help predict the deterioration risk in patients with normotensive acute pulmonary embolism. REGISTRATION: https://clinicaltrials.gov; identifier: NCT05098769.

Environments of tornadic and non‐tornadic supercells in China and optimized significant tornado parameter for China region

Abstract With a sample of 273 supercells spanning 20 years, inflow environment differences between tornadic and non‐tornadic supercells in China and its three subregions (northern, central and southern China; CNN, CNC, and CNS) are examined using sounding‐derived parameters. Proximity soundings are extracted from the hourly ERA5 reanalysis data. The supercells are categorized as significantly tornadic [rated (E)F2+], weakly tornadic [rated (E)F1], and non‐tornadic. Thermodynamic parameters, such as convective available potential energy (CAPE), lifting condensation level (LCL), low‐level relative humidity (RH) and convective inhibition (CIN), cannot discriminate between tornadic and non‐tornadic supercells effectively. In addition, thermodynamic parameters based on mixed‐layer (ML) lifted parcels show worse skill than those for surface‐based (SB) or most unstable (MU) lifted parcels. Storm‐relative helicity (SRH300) in the range 0–300 m and 0–300‐m bulk shear (SHR300) demonstrate greater forecasting skills compared to SRH and shear over deeper depths. Based on predictive skills and distributions of individual parameters, a new significant tornado parameter (STP) formulation, STP300cn, using MUCAPE, MULCL, MUCIN, SRH300, and SHR300 is composed. True skill score (TSS) is used to measure the capability of the individual or combined parameters in discriminating significantly tornadic from non‐tornadic supercells. The thresholds and normalization factors for terms in STP are calibrated to the China cases to obtain optimal predictive TSS scores. The calibrated STP parameter, called STP300cn, achieves a TSS of 0.51 in China overall, compared to the 0.14 and 0.29 of the two original versions of STP. It achieves a TSS of 0.37, 0.66, 0.42 for CNN, CNC and CNS, respectively, all much higher than those of the original STP parameters.

Evaluating the Performance of In silico Tools for PRRT2 Missense Variants.

Variants in the PRRT2 gene are associated with paroxysmal kinesigenic dyskinesia and other episodic disorders. With the employment of variant screening in patients with episodic dyskinesia, many PRRT2 variants have been discovered. Bioinformatics tools are becoming increasingly important for predicting the functional significance of variants. This study aimed to evaluate the performance of six in silico tools for PRRT2 missense variants. Pathogenic PRRT2 variants were retrieved from the Human Gene Mutation Database (HGMD) and literature from the PubMed database. The benign set of non-deleterious variants was retrieved from the Genome Aggregation Database (gnomAD). The overall accuracy, sensitivity, specificity, positive predictive values, and negative predictive values of SIFT, PolyPhen2, MutationTaster, CADD, Fathmm, and Provean were analyzed. The MCC score and ROC curve were calculated. The GraphPad Prism 8.0 software was used to plot ROC curves for the six bioinformatics software. A total of 45 missense variants with confirmed pathogenicity were used as a positive set, and 222 missense variants were used as a negative set. The top three tools in accuracy are Fathmm, Provean, and MutationTaster. The top three predictors in sensitivity are SIFT, PolyPhen2, and CADD. Regarding specificity, the top three tools were Provean, Fathmm, and MutationTaster. In terms of the MCC and F-score, the highest degree was observed in Fathmm. Fathmm also had the highest AUC score. The cutoff values of Fathmm, CADD, PolyPhen2, and Provean were between the median prediction scores of the positive and negative sets. In contrast, the cutoff value of SIFT was below the median prediction score of the positive and negative sets. Fathmm had the highest accuracy. The prediction performance of six in silico tools differed among the parameters. Fathmm had the best prediction performance, with the highest accuracy and MCC/F-score for PRRT2 missense variants.

Validation of a computed tomography-based model for early prediction of postpancreatectomy hemorrhage risk.

Validation of a computed tomography-based model for early prediction of postpancreatectomy hemorrhage risk.

Model of End-Stage Liver Disease-alpha-fetoprotein-tumor burden (MELD-AFP-TBS) score to stratify prognosis after liver resection for hepatocellular carcinoma.

Model of End-Stage Liver Disease-alpha-fetoprotein-tumor burden (MELD-AFP-TBS) score to stratify prognosis after liver resection for hepatocellular carcinoma.

Predictability and Familiarity Assessments for Greek Idiomatic Expressions: the Role of Reading Habits and Language Profiles

This study examines the predictability and familiarity ratings of 199 Greek idiomatic expressions, building on the random sample used by Lada et al. (2024). The primary goals are to explore correlations between predictability, familiarity, and idiom dimensions such as decomposability, subjective frequency, and ambiguity, and to investigate the relationships between idiom familiarity, predictability, and participants' bilingual/multilingual profiles and reading habits. Sixty-three native Greek-speaking students at Democritus University of Thrace completed familiarity and predictability assessments based on a random selection of idioms from Vlaxopoulos (2007). Correlational analyses, aligned with Lada et al. (2024), show that subjective frequency is positively correlated with both ambiguity and decomposability. In addition, familiarity is weakly correlated with ambiguity, moderately correlated with decomposability, and strongly correlated with subjective frequency and predictability. Furthermore, predictability is weakly correlated with ambiguity, moderately correlated with subjective frequency and decomposability, but strongly correlated with familiarity. Logistic regression analyses reveal that the number of foreign languages spoken negatively predicts correct idiom completion in the predictability task, with more languages associated with lower predictability scores. Mixed-effects linear models indicate that higher reading frequency is linked to lower familiarity ratings, whereas more books read is associated with higher familiarity. These findings provide novel insights into idiom comprehension among bilinguals, highlighting the influence of language profiles and reading habits on idiom familiarity and predictability. Limitations include the binary approach to predictability scoring and the lack of language-specific details. We hence suggest future studies consider typological factors and alternative results’ interpretation for idiom predictability.

External multicenter validation of the FAMISH score for predicting metachronous gastric lesions after endoscopic submucosal dissection

Background and objectives Surveillance after gastric endoscopic submucosal dissection (ESD) is crucial due to the high risk of metachronous gastric lesions (MGL), although this risk may differ between patients. We sought to validate the FAMISH score – a prediction score to estimate MGL risk after gastric ESD – within a multicentric framework. Materials and methods Performance measures of the FAMISH score were assessed in a retrospective analysis of a multicenter cohort, which included consecutive adult patients undergoing ESD for a primary gastric superficial lesion at 15 international centers, with a minimum endoscopic follow-up of at least 3 years. Results A total of 855 individuals were included, with 20% of them considered low-risk according to the FAMISH score. After a mean follow-up time of 5 years (SD ± 2 years), 168 patients (19.6%) developed MGL. At 3 years of follow-up, the score achieved 90.4% sensitivity and 93.9% NPV. At 5 years follow-up, the score achieved 89.1% sensitivity and 85.3% NPV. The FAMISH risk score achieved a fair diagnostic accuracy with an AUC of 0.618 at 3 years (p < 0.001) and 0.597 (p = 0.006) at 5 years of follow-up. The progression to MGL at 5 years of follow-up was significantly lower for the low-risk group (9.6% vs. 18.2%, p = 0.029). Conclusions The FAMISH score achieved an acceptable diagnostic accuracy in a multicentric validation cohort from Western countries. This score is a useful tool to identify patients with low risk for MGL allowing to safely extend surveillance intervals and reduce the burden of care.

Exploratory Analysis of Practical Predictive Indices for the Efficacy of Mogamulizumab in Patients With Aggressive Adult T‐Cell Leukemia‐Lymphoma

ABSTRACTAn exploratory analysis of past clinical trials was conducted to propose a predictive scoring system for the efficacy of mogamulizumab, an anti‐CC chemokine receptor 4 (CCR4) antibody, based on easily measurable parameters. Factors affecting progression‐free survival (PFS) were investigated using data from three clinical trials (NCT00920790, NCT01626664, and NCT01173887) and one clinical study (UMIN000013294) conducted in patients with relapsed/refractory (R/R) or untreated CCR4‐positive aggressive adult T‐cell leukemia‐lymphoma (ATL) receiving mogamulizumab treatment. Twelve routinely measured clinical parameters and three calculated indices—lymphocyte‐to‐neutrophil count ratio, platelet‐to‐lymphocyte count ratio, and lymphocyte‐to‐monocyte count ratio (LMR)—were selected as variables. Univariate Cox proportional hazards analysis identified albumin level, disease type, lactate dehydrogenase (LDH), monocyte count, neutrophil count, and LMR as relevant factors in R/R ATL patients treated with mogamulizumab monotherapy (p < 0.05). A predictive model constructed from multivariate analysis results stratified the monotherapy group (n = 69) into three subgroups, with scores of 0 (n = 5), 1 (n = 25), and 2 (n = 39), based on LDH (0 for < 265 and 1 for ≥ 265) and LMR (0 for ≥ 3.571 and 1 for < 3.571). Median PFS values were 0.57, 0.46, and 0.07 years for scores 0, 1, and 2, respectively (log‐rank test: p = 0.005 for score 0 vs. 2; p < 0.001 for score 1 vs. 2). The simple model combining LDH and LMR may predict PFS in patients with R/R aggressive ATL receiving mogamulizumab treatment. Since LDH and LMR are easily measurable in clinical practice, this model could help predict mogamulizumab efficacy and guide treatment decisions in this patient population.Trial Registration: Registration number: UMIN000049135. Date of registration: October 17, 2022

Radiomic 'Stress Test': exploration of a deep learning radiomic model in a high-risk prospective lung nodule cohort.

Indeterminate pulmonary nodules (IPNs) are commonly biopsied to ascertain a diagnosis of lung cancer, but many are ultimately benign. The Lung Cancer Prediction (LCP) score is a commercially available deep learning radiomic model with strong diagnostic performance in incidentally identified IPNs, but its potential use to reduce the need for invasive procedures has not been evaluated in patients with nodules for which a biopsy has been recommended. In this prospectively collected, retrospective blinded evaluation, the probability of cancer in consecutively biopsied IPNs at a tertiary care centre was calculated using the Mayo Clinic prediction model and categorised into low, intermediate and high-probability groups by applying <10% no-test and >70% treatment thresholds per British Thoracic Society guidelines. We evaluated the diagnostic performance of the Mayo Clinic model, the LCP radiomic model and an integrated model combining the LCP score with statistically selected clinical variables (age, spiculation and upper lobe location) using stepwise logistic regression. Performance was assessed using area under the receiver operating characteristic curve (AUC), F1 score and reclassification analysis based on the bias-corrected clinical net reclassification index. The study population included 196 malignant and 125 benign IPNs (61% prevalence of malignancy). The Mayo Clinic model's AUC was 0.69 (0.63-0.75), LCP's AUC was 0.67 (0.61-0.73) and the integrated model combining LCP with statistically selected clinical variables (age, spiculation and upper lobe location) had the highest AUC of 0.75 (0.69-0.80). The integrated model demonstrated improved classification, with an F1 score of 0.645 (0.572-0.716) and a significantly higher AUC compared with the Mayo Clinic model (p=0.046). Reclassification analysis showed a clinical net reclassification index of 0.36 (0.21-0.53) for benign IPNs with eight correctly downgraded intermediate-risk benign nodules and no malignant nodules misclassified into the low-risk category. Incorporating LCP with select clinical variables results in an improvement in malignancy risk prediction and nodule classification and could reduce unnecessary invasive biopsies for IPNs.

Use of Preoperative Renal Volumetry to Predict Postdonation Renal Function With Inulin Clearance as Indicator in Japan.

Use of Preoperative Renal Volumetry to Predict Postdonation Renal Function With Inulin Clearance as Indicator in Japan.

Revisiting Essential and Nonessential Settings of Evidential Deep Learning.

Evidential Deep Learning (EDL) is an emerging method for uncertainty estimation that provides reliable predictive uncertainty in a single forward pass, attracting significant attention. Grounded in subjective logic, EDL derives Dirichlet concentration parameters from neural networks to construct a Dirichlet probability density function (PDF), modeling the distribution of class probabilities. Despite its success, EDL incorporates several nonessential settings: In model construction, (1)a commonly ignored prior weight parameter is fixed to the number of classes, while its value actually impacts the balance between the proportion of evidence and its magnitude in deriving predictive scores. In model optimization, (2)the empirical risk features a variance-minimizing optimization term that biases the PDF towards a Dirac delta function, potentially exacerbating overconfidence. (3)Additionally, the structural risk typically includes a KL-divergence-minimizing regularization, whose optimization direction extends beyond the intended purpose and contradicts common sense, diminishing the information carried by the evidence magnitude. Therefore, we propose Re-EDL, a simplified yet more effective variant of EDL, by relaxing the nonessential settings and retaining the essential one, namely, the adoption of projected probability from subjective logic. Specifically, Re-EDL treats the prior weight as an adjustable hyperparameter rather than a fixed scalar, and directly optimizes the expectation of the Dirichlet PDF provided by deprecating both the variance-minimizing optimization term and the divergence regularization term. Extensive experiments and state-of-the-art performance validate the effectiveness of our method. The source code is available at https://github.com/MengyuanChen21/Re-EDL.

Role of elastography and dynamic contrast-enhanced ultrasound in the evaluation of pancreas transplantation rejection.

Rejection is the leading cause of graft failure, and its diagnosis remains a challenge. Elastography and dynamic contrast-enhanced ultrasound (DCE-US) are novel non-invasive techniques for quantifying tissue elasticity and perfusion. Their role in pancreas graft rejection has not yet been defined. From October 16 to January 20, all pancreas transplantations performed at our institution were prospectively studied with elastography and DCE-US at 1 week, 3 weeks, and 12 months post-transplantation. Surveillance biopsies were performed at 3 weeks and 12 months. Elastography and DCE-US were also conducted in all requested biopsies during this period (regardless of the date of transplantation). Patients were categorized according to the biopsy result: normal/rejection. Grafts with other complications were excluded. Cut-off values were established. One hundred twenty-one elastography and 127 DCE-US in 56 patients were included. All parameters showed a high dispersion during the first 90 days post-transplantation. After this period, the rejection group presented higher stiffness (0.97 vs 1.46 m/s, p < 0.001) and lower perfusion. The optimal cut-off value for elastography was 1.27 m/s (AUC 0.80), and for DCE-US were: peak enhancement 601 a.u. (AUC 0.67), wash-in AUC 2748 a.u. (AUC 0.70), wash-in rate 118 a.u. (AUC 0.65), wash-in perfusion index 369 a.u. (AUC 0.67), wash-out AUC 5181 a.u. (AUC 0.69) and total AUC 6388 a.u. (AUC 0.68). A combined predictive score showed that alteration of elastography and DCE-US was associated with a 23.2-fold probability of rejection. After the first 90 days post-transplantation, pancreas graft rejection is associated with higher stiffness and lower graft perfusion. Question Pancreas graft rejection remains a clinical challenge, as there are currently no reliable non-invasive tests for its diagnosis. Findings After the first 90 days post-transplantation, elastography and DCE-US show higher stiffness and lower pancreas graft perfusion in the presence of rejection. Clinical relevance These non-invasive tools, which can be easily integrated into daily routine practice, may be useful in identifying grafts at higher risk of rejection, allowing closer follow-up or early biopsy to establish early rejection treatment, improving graft and patient survival.

IPSS-M risk and specific sex-associated somatic mutations predict response to ESA therapy in LR-MDS: building a new score.

IPSS-M risk and specific sex-associated somatic mutations predict response to ESA therapy in LR-MDS: building a new score.

Optimizing Vanadium-Catalyzed Epoxidation Reactions: Machine-Learning-Driven Yield Predictions and Data Augmentation.

Catalytic epoxidations are key chemical processes serving as essential steps in the synthesis of commercially valuable compounds. This study presents an innovative supervised machine learning (ML) model to predict the reaction yield of the vanadium-catalyzed epoxidation of small alcohols and alkenes. Our framework uncovers relevant chemical characteristics for structure design, offering a pathway for automated optimization of epoxidation reactions. The study also incorporates the concept of data augmentation, handling experimental variability by generating synthetic reactions to densify under-represented data segments. Trained on a curated data set of 273 experimental epoxidation reactions with vanadyl catalyst groups, the model achieved a predictive R2 test score of 90%, with a mean absolute yield prediction error of 4.7%. The ML model offers a high degree of explainability, as descriptor analysis identified key experimental and chemical descriptors that influence catalytic reaction predictions. This represents a significant development in catalytic epoxidation studies, highlighting the critical role of data science in reaction research and catalyst optimization.