Liqing Li

Background: Infections are a major cause of mortality in hemodialysis patients, but early risk stratification remains difficult. This study developed and validated an interpretable ML model using routine baseline laboratory data to predict short-term infection risk. Methods: A retrospective cohort of 622 hemodialysis patients with uremia was analyzed. The primary outcome was a confirmed infectious event (infection, sepsis, or pneumonia) occurring within a 2-month follow-up. To rigorously prevent data leakage, the cohort was strictly partitioned into training (70%) and test (30%) sets prior to missing value imputation and any downstream preprocessing. A LightGBM-based forward feature selection was executed exclusively within the training set. Nine ML algorithms were evaluated, and model robustness was validated via nested cross-validation, repeated cross-validation, and bootstrap optimism correction. Interpretability was achieved using SHapley Additive exPlanations (SHAP). Results: Subsequent infections occurred in 157 patients (25.2%). A parsimonious six-biomarker signature (CysC, HCT, TP, MONO_pct, MYO, ADA) was identified. The K-Nearest Neighbors (KNN) and LightGBM models demonstrated robust discriminative performance (Test AUROC: 0.727 and 0.719, respectively). Nested cross-validation confirmed generalization stability (e.g., ANN mean AUC: 0.722). SHAP analysis revealed elevated Cystatin C and lowered Hematocrit as the strongest contributors to the model’s risk ranking. Although the model showed high ranking ability, the default 0.5 threshold yielded low sensitivity; a prespecified F2‑optimized threshold of 0.30 improved sensitivity to 0.723 in the test set. Conclusion: This interpretable machine learning‑based predictive model, using routinely available laboratory data, may serve as a potential adjunctive decision‑support tool for stratifying short‑term infection risk. External validation is required before clinical implementation.

Purpose: Patients with hematological malignancies who develop sepsis exhibit profound immune suppression and atypical inflammatory responses, yet the longitudinal dynamics of the neutrophil‑to‑lymphocyte ratio (NLR) in this population remain poorly characterized. We aimed to identify reproducible NLR trajectory archetypes in hematological sepsis, externally validate them, and evaluate their added prognostic value over static severity scoring. Methods: We analyzed 2,131 patients with hematological sepsis from MIMIC‑IV (derivation) and 930 from eICU and NWICU (external validation). A competing‑risk latent class mixed model was fitted to longitudinal NLR measurements, accounting for informative censoring due to death and ICU discharge. The primary endpoint was 14‑day mortality. XGBoost classifiers predicted trajectory membership by day 2 using 12 routine clinical variables. Added prognostic value was assessed via DeLong's test, continuous Net Reclassification Improvement (cNRI), and Decision Curve Analysis. Trajectories were mapped to SENECA sepsis subphenotypes. Results: A three‑class model provided optimal fit. Class 1 (Ascending, 57.9%) had the lowest 14‑day mortality (13.5%). Class 2 (Descending, 5.5%) exhibited an extremely elevated baseline NLR (median 107.9) with the highest 14‑day mortality (27.8%). Class 3 (Stable, 36.6%) showed a 14‑day mortality of 24.9%. Archetypes replicated in both external cohorts. Class 2 was overwhelmingly dominated by the hyperinflammatory Delta phenotype (82.6% in MIMIC‑IV, 76.5% in eICU, 75.0% in NWICU). By day 2, the interaction‑enhanced trajectory model achieved an AUC of 0.744 versus 0.719 for the static baseline (DeLong *p* = 0.025), with a striking cNRI of 0.574 and superior net clinical benefit. Conclusions: Three reproducible NLR archetypes capture distinct immunodynamic profiles in hematological sepsis. The interaction‑enhanced trajectory model substantially improved risk reclassification, offering a practical tool for early risk stratification in this highly immunocompromised population.

Background: Prostate cancer (PCa) is characterized by pronounced intratumoral heterogeneity and multifocality, presenting ongoing challenges for early and precise diagnosis. In clinical practice, a substantial proportion of patients initially diagnosed with benign prostatic hyperplasia (BPH) are subsequently found to harbor latent PCa during follow-up. However, current prostate-specific antigen (PSA)-based screening lacks the specificity required to resolve this diagnostic overlap. This study aimed to develop and internally validate an interpretable machine learning (ML) model as an adjunct to PSA testing for identifying latent PCa in patients initially diagnosed with BPH. Methods: This retrospective cohort study utilized data from 1,071 patients to develop a machine learning framework aimed at identifying pre-existing occult PCa diagnosed at least 6 months after the initial BPH assessment. The cohort was partitioned into training (70%) and test (30%) subsets. Within the training set, Recursive Feature Elimination (RFE) was utilized to identify a concise biomarker signature. Six machine learning algorithms were evaluated: logistic regression (LR), random forest (RF), extreme gradient boosting (XGB), support vector machine (SVM), k-nearest neighbors (KNN), and multilayer perceptron (MLP). Model performance was assessed using the area under the receiver operating characteristic curve (AUROC), Brier score, and decision curve analysis (DCA). Model interpretability was facilitated through SHapley Additive exPlanations (SHAP) and Locally Weighted Scatterplot Smoothing (LOWESS) analysis. Results: A concise four-biomarker signature comprising total prostate-specific antigen (tPSA), lactate dehydrogenase (LDH), red blood cell count (RBC), and creatine kinase-MB (CK-MB) was identified. The MLP and LR models exhibited high diagnostic performance, with test AUROC values of 0.949 and 0.946, respectively. Conclusion: This research developed and internally validated an interpretable machine learning framework designed to complement traditional PSA screening methodologies. The tool is intended exclusively for research purposes.

The integration of immunotherapy with neoadjuvant therapy for proficient mismatch repair (pMMR) locally advanced rectal cancer (LARC) is a promising strate

Objective: To develop an interpretable machine learning (ML) model using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) radiomic data, dosimetric parameters, and clinical data for predicting radiation-induced hepatic toxicity (RIHT) in patients with hepatocellular carcinoma (HCC) following intensity-modulated radiation therapy (IMRT). Methods: A retrospective analysis of 150 HCC patients was performed, with a 7:3 ratio used to divide the data into training and validation cohorts. Radiomic features from the original MRI sequences and Delta-radiomic features were extracted. Seven ML models based on radiomics were developed: logistic regression (LR), random forest (RF), support vector machine (SVM), eXtreme Gradient Boosting (XGBoost), adaptive boosting (AdaBoost), decision tree (DT), and artificial neural network (ANN). The predictive performance of the models was evaluated using receiver operating characteristic (ROC) curve analysis and calibration curves. Shapley additive explanations (SHAP) were employed to interpret the contribution of each variable and its risk threshold. Results: Original radiomic features and Delta-radiomic features were extracted from DCE-MRI images and filtered to generate Radiomics-scores and Delta-Radiomics-scores. These were then combined with independent risk factors (Body Mass Index (BMI), V5, and pre-Child-Pugh score(pre-CP)) identified through univariate and multivariate logistic regression and Spearman correlation analysis to construct the ML models. In the training cohort, the AUC values were 0.8651 for LR, 0.7004 for RF, 0.6349 for SVM, 0.6706 for XGBoost, 0.7341 for AdaBoost, 0.6806 for Decision Tree, and 0.6786 for ANN. The corresponding accuracies were 84.4%, 65.6%, 75.0%, 65.6%, 71.9%, 68.8%, and 71.9%, respectively. The validation cohort further confirmed the superiority of the LR model, which was selected as the optimal model. SHAP analysis revealed that Delta-radiomics made a substantial positive contribution to the model. Conclusion: The interpretable ML model based on radiomics provides a non-invasive tool for predicting RIHT in patients with HCC, demonstrating satisfactory discriminative performance.

This study aimed to construct a nomogram to predict radiation-induced hepatic toxicity in patients with hepatocellular carcinoma treated with intensity-modulated radiotherapy. This study reviewed the clinical characteristics and dose–volume parameters of 196 patients with hepatocellular carcinoma. Radiation-induced hepatic toxicity was defined as progression of the Child–Pugh score caused by intensity-modulated radiotherapy. Factors relevant to radiation-induced hepatic toxicity were selected using receiver operating characteristic and univariate logistic analysis. A risk assessment model was developed, and its discrimination was validated. Eighty-eight (44.90%) and 28 (14.29%) patients had radiation-induced hepatic toxicity ≥ 1 (Child–Pugh ≥ 1) and radiation-induced hepatic toxicity ≥ 2 (Child–Pugh ≥ 2). Pre-treatment Child–Pugh, body mass index and dose–volume parameters were correlated with radiation-induced hepatic toxicity ≥ 1 using univariate logistic analysis. V15 had the best predictive effectiveness among the dose–volume parameters in both the training (area under the curve: 0.763, 95% confidence interval: 0.683–0.842, P < 0.001) and validation cohorts (area under the curve: 0.759, 95% confidence interval: 0.635–0.883, P < 0.001). The area under the curve values of the model that was constructed by pre-treatment Child–Pugh, body mass index and V15 for radiation-induced hepatic toxicity ≥1 were 0.799 (95% confidence interval: 0.719–0.878, P < 0.001) and 0.775 (95% confidence interval: 0.657–0.894, P < 0.001) in the training and validation cohorts, respectively. Patients with a body mass index ≤ 20.425, Barcelona clinic liver cancer = C, Hepatitis B Virus-positive, Eastern Cooperative Oncology Group = 1–2 and hepatic fibrosis require lower V15 dose limits. Risk assessment model constructed from Pre-treatment Child–Pugh, V15 and body mass index can guide individualized patient selection of toxicity minimization strategies.

Computed tomography (CT)–assessed body composition is considered a novel prognostic factor for cancer patients. Owing to the need for new prognostic markers for hepatocellular carcinoma (HCC) patients undergoing radiotherapy, we investigated the impact of body composition on outcomes in this patient population. We retrospectively evaluated 109 HCC patients receiving radiotherapy. The skeletal muscle index, subcutaneous adipose tissue index (SATI), and visceral adipose tissue index within 1 mo, before radiotherapy were assessed based on a single CT image slice at the level of the third lumbar (L3) vertebra. The impact of body composition parameters on progression-free survival (PFS) and overall survival (OS) was assessed. Overall, 62 (56.9%) patients died, and 47 (43.1%) patients experienced recurrence during a median follow-up period of 20.5 mo. Multivariate analysis revealed that SATI was an independent prognostic factor for both PFS (hazard ratio [HR] 0.542, P = 0.025) and OS (HR 0.385, P = 0.005). Patients with high SATI (n = 43) had significantly better PFS (P = 0.0093) and OS (P = 0.032) than those with low SATI (n = 66). CT-assessed SATI is an independent prognostic factor in HCC patients receiving radiotherapy. Further validation is warranted to determine whether this finding can be translated into other study populations.

Background It is unclear whether robotic stereotactic body radiotherapy (SBRT) is superior to intensity-modulated radiotherapy (IMRT) in advanced hepatocellular carcinoma (HCC). This study aimed to compare the long-term outcomes of SBRT with those of IMRT in HCCs with portal vein tumor thrombosis (PVTT). Methods We retrospectively evaluated 287 HCC patients with PVTT who underwent radiotherapy between January 2000 and January 2017. Of them, 154 and 133 patients were treated with IMRT and SBRT, respectively. Overall survival (OS), progression-free survival (PFS), intrahepatic control (IC), and local control (LC) were evaluated in univariable and propensity-score matched analyses. Results After matching, 102 well-paired patients were selected. There was no significant difference in the 6-, 12-, 24-, and 60-month cumulative OS (73.5, 42.9, 23.6, 7.6% vs. 72.4, 45.1, 29.8, 13.2%, p = 0.151), PFS (53.9, 29.3, 21.8, 7.5% vs. 54.5, 19.3, 12.0, 9.6%, p = 0.744), IC (61.4, 45.7, 39.0, 26.8% vs. 75.1, 45.8, 35.9, 28.7%, p = 0.144), and LC (85.2, 56.5, 52.1, 47.4% vs. 87.4, 65.2, 62.1, 62.1%, p = 0.191) between the IMRT and SBRT groups. A biologically effective dose assumed at an a/b ratio of 10 (BED₁₀) of ≥ 100 Gy was the optimal cutoff for predicting the OS, PFS, IC, and LC in the patients who received SBRT. Conclusions When high-precision tracking technology is available, SBRT appears to be a safe and more time-efficient treatment, achieving comparable OS, PFS, IC and LC to IMRT for local advanced HCC with PVTT. A BED₁₀ ≥ 100 Gy is recommended if tolerated by normal tissue. Graphic abstract

Background: It is unclear whether robotic stereotactic body radiotherapy (SBRT) is superior to intensity-modulated radiotherapy (IMRT) in advanced hepatocellular carcinoma (HCC). This study aimed to compare the long-term outcomes of SBRT with those of IMRT in HCCs with portal vein tumor thrombosis (PVTT). Methods: We retrospectively evaluated 287 HCC patients with PVTT who underwent radiotherapy between January 2000 and January 2017. Of them, 154 and 133 patients were treated with IMRT and SBRT, respectively. Overall survival (OS), progression-free survival (PFS), intrahepatic control (IC), and local control (LC) were evaluated in univariable and propensity-score matched analyses. Results: After matching, 102 well-paired patients were selected. There was no significant difference in the 6-, 12-, 24-, and 60-month cumulative OS (73.5, 42.9, 23.6, 7.6% vs. 72.4, 45.1, 29.8, 13.2%, P=0.151), PFS (53.9, 29.3, 21.8, 7.5% vs. 54.5, 19.3, 12.0, 9.6%, P=0.744) , IC (61.4, 45.7, 39.0, 26.8% vs. 75.1, 45.8, 35.9, 28.7%, P=0.144), and LC (85.2, 56.5, 52.1, 47.4% vs. 87.4, 65.2, 62.1, 62.1%, P=0.191) between the IMRT and SBRT groups. A biologically effective dose assumed at an a/b ratio of 10 (BED₁₀) of ≥100 Gy was the optimal cutoff for predicting the OS, PFS, IC, and LC in the patients who received SBRT. Conclusions: When high-precision tracking technology is available, SBRT appears to be a safe and more time-efficient treatment, achieving comparable OS, PFS, IC and LC to IMRT for local advanced HCC with PVTT. A BED₁₀≥100 Gy is recommended if tolerated by normal tissue.