Aug 2026· Frontiers in Surgery· 0 citations· 25 references
TL;DR
The LASSO-based GBM model demonstrated stable predictive performance and acceptable clinical utility and was selected as the final model for predicting postoperative recurrence of anal fistula.
Abstract
This study aimed to develop and interpret a machine learning model for predicting postoperative recurrence of anal fistula using routine laboratory indicators and inflammation-related indices.
A total of 2,214 patients who underwent fistulectomy were included. Patients from wards 5, 11, 12, 13 and 14 (
n
= 1,772) were divided by stratified random sampling according to recurrence status into training (
n
= 1,242) and testing (
n
= 530) cohorts. Patients from wards 15 and 16 (
n
= 442), which were managed by separate clinical teams, were reserved as a ward-based internal validation cohort. Univariate and multiple fistula tracts. analyses were performed to identify recurrence-associated factors, and LASSO regression was used for feature selection. Multiple machine learning models were developed and compared, including logistic regression, support vector machine, GBM, neural network, XGBoost, AdaBoost, LightGBM, and CatBoost. Model performance was assessed using ROC curves, calibration curves, decision curve analysis, and classification metrics. SHAP analysis was applied for model interpretation.
Multivariate logistic regression analysis showed that WBC, RBC, hs-CRP, and NCR were independent predictors of recurrence. LASSO regression selected 11 variables for model development. Among the candidate models, GBM demonstrated the most balanced predictive performance and was therefore selected as the final model. The AUCs of GBM in the training, testing, and validation sets were 0.777, 0.784, and 0.712, respectively. Calibration curves showed acceptable agreement between predicted and observed risks, while decision curve analysis indicated potential clinical benefit within low-to-moderate threshold probability ranges. SHAP analysis identified age, WBC, RBC, NCR, and hs-CRP as the main contributors to model prediction. Restricted cubic spline analysis revealed a significant nonlinear association between NCR and recurrence risk.
WBC, RBC, hs-CRP, and NCR were independently associated with postoperative recurrence of anal fistula. The LASSO-based GBM model demonstrated stable predictive performance and acceptable clinical utility. Routine hematological parameters and inflammation-related indices, particularly NCR, may support individualized recurrence risk stratification and postoperative follow-up.
Background This study aimed to develop, validate and evaluate interpretable machine learning models using clinical and laboratory data for prognosis prediction in patients with primary biliary cholangitis (PBC). Methods This study included 7905 patients treated for PBC. The cohort data were randomly divided into training and testing sets, with external validation using an additional 2372 patients. Six machine learning models are compared in this research (logistic regression (LR), random forest (RF),Extreme Random Trees (ET), Extreme Gradient Boosting (XGBoost), Lightweight Gradient Boosting Machine (LightGBM), and Multi-Layer Perceptron (MLP)). Feature importance and model interpretation were analyzed using the SHapley Additive exPlanations (SHAP) method. Results Ensemble tree models demonstrated significantly superior performance compared to traditional linear models and shallow neural networks in predicting PBC patient prognosis. Among these, the Extreme Random Tree model exhibited optimal predictive efficacy on both the training set (AUC = 0.9898) and external validation set (AUC = 0.9681), while the Random Forest model showed comparable performance with greater stability. At the patient level, SHAP’s force maps and decision trees provided clinically meaningful explanations for the et algorithm. The bilirubin_albumin_ratio emerged as the core feature for predicting PBC prognosis, with bilirubin, n_days, and prothrombin serving as key influencing factors. The influence patterns of these features align closely with clinical and pathological mechanisms. Conclusion The ET model constructed in this study enables precise prognosis prediction for PBC patients. After SHAP analysis, it demonstrates good interpretability. The key prognostic features identified by the model provide quantitative evidence for clinically assessing disease severity in PBC patients and offer data support for developing individualized clinical intervention plans.
Yifeng Dou, Jiantao Liu· Frontiers in Molecular Biosc...· 0 citations
AIMS/BACKGROUND
Post-chemotherapy infection is a major cause of treatment failure in lung cancer (LC) patients undergoing neoadjuvant chemotherapy (NAC). This study aimed to identify risk factors for post-NAC infection using the Least Absolute Shrinkage and Selection Operator (LASSO) regression and to develop and validate a corresponding risk prediction model.
METHODS
Clinical data from 144 LC patients who underwent NAC at Hunan Provincial People's Hospital between January 2021 and December 2024 were retrospectively analysed. Patients were stratified into a non-infection group (n = 81) and an infection group (n = 63) based on the occurrence of infection. LASSO-logistic regression was used to identify risk factors for post-chemotherapy infection. A risk prediction nomogram was subsequently constructed and validated based on these factors.
RESULTS
The optimal LASSO model was selected at lambda.1se (λ = 0.074), which retained 9 of the 15 candidate predictors. Eastern Cooperative Oncology Group Performance Status (ECOG-PS) ≥2, recent invasive procedures/catheterization, and Nutritional Risk Screening 2002 (NRS-2002) score ≥3 were identified as significant risk factors for post-chemotherapy infection, while a high cluster of differentiation 4-positive (CD4+) count served as a protective factor (p < 0.05). A nomogram incorporating these variables was subsequently developed. Internal validation using the bootstrap method (1000 iterations) demonstrated a good predictive performance with an area under the curve (AUC) of 0.831 (95% confidence interval [CI]: 0.761-0.901, p < 0.001), sensitivity of 76.2%, and specificity of 79.0% at the optimal cutoff. The calibration curve demonstrated good agreement between predicted and observed outcomes. Decision curve analysis (DCA) confirmed the clinical utility of the nomogram by showing a positive net benefit across a wide range of threshold probabilities.
CONCLUSION
The LASSO-derived nomogram integrates key variables (ECOG-PS ≥2, invasive procedures/catheterization, NRS-2002 score ≥3, and CD4+ level) to enable individualised prediction of post-NAC infection risk in LC patients. Targeted interventions addressing these risk factors, together with maintenance of protective factors, may help reduce infection rates and improve treatment outcomes, providing a scientific basis for infection prevention and management.
Unknown authors· British journal of hospital...· 0 citations
Objective To investigate prognostic factors for breast cancer recurrence and metastasis, and to systematically compare multiple machine learning models to develop an optimal predictive tool. Methods We retrospectively analyzed data from 1,056 breast cancer patients diagnosed between January 2012 and June 2021 at a single center. Patients were randomly divided into a training (n=740) and a validation (n=316) set. Univariate and multivariate Cox proportional hazards regression analyses were performed to identify independent prognostic factors. Seven machine learning algorithms (Cox regression, LASSO, Elastic-Net, Decision Tree, Random Forest, XGBoost, and GBM) were employed. All models were implemented using survival-specific adaptations. A rigorous 5-fold cross-validation framework was used for model training and hyperparameter tuning. Model performance was evaluated using time-dependent Area Under the Curve (AUC), Brier scores, calibration curves, calibration-in-the-large, calibration slopes, and Decision Curve Analysis (DCA). SHAP values were employed for model interpretation. Results Multivariate Cox regression revealed that tumor size (cm) (HR = 1.025, 95%CI: 1.013-1.037), lymph node dissection (HR = 0.278, 95%CI: 0.199-0.389), ER% (HR = 1.006, 95%CI: 1.003-1.009), PR% (HR = 1.005, 95%CI: 1.002-1.008), Ki-67% (HR = 1.012, 95%CI: 1.007-1.016), and HER2 status (HR = 1.195, 95%CI: 1.098-1.301) were independently associated with disease-free survival. Random Forest and XGBoost demonstrated superior and stable predictive performance, with Random Forest achieving time-dependent AUCs of 0.851 (95%CI: 0.802-0.900, 1-year), 0.763 (95%CI: 0.702-0.824, 3-year), and 0.826 (95%CI: 0.766-0.886, 5-year) in the validation set. The Brier scores for Random Forest were consistently low, and calibration metrics confirmed excellent calibration. DCA indicated a positive net benefit across a wide range of threshold probabilities. Conclusion This single-center study identifies key prognostic factors for breast cancer and demonstrates that ensemble machine learning models, particularly Random Forest, offer superior predictive power. The integration of SHAP interpretation provides a methodological framework for potential clinical application. However, all predictive models require external, multi-center validation before clinical consideration. These findings provide a promising methodological basis, but caution is warranted against overinterpretation until independent verification is complete.
Meiying Shen, Yulei Wang, Zong-ming Wu et al.· Frontiers in Oncology· 0 citations
Esophageal Cancer: Other
Esophagectomy is associated with substantial morbidity, mortality, and resource utilization. Traditional regression-based risk tools may inadequately capture complex nonlinear interactions. Contemporary evidence on machine learning (ML) models predicting postoperative outcomes after esophagectomy was synthesized, focusing on discrimination, validation, and comparison with conventional regression approaches.
A PRISMA-guided systematic review was conducted using Embase, MEDLINE, PubMed, and the Cochrane Library in January 2026. A total of 196 studies were identified. After title and abstract screening, 32 studies underwent full-text review, of which 10 met final inclusion criteria as ML-focused prediction models in esophagectomy populations. Extracted data included study design, cohort size, procedure type, predicted outcome, modeling approach (ML versus regression), validation strategy (internal or external), performance metrics (e.g., area under the receiver operating characteristic curve [AUROC]), and reporting elements such as calibration and decision-curve analysis. ML-focused studies were defined as those applying algorithms including gradient boosting, support vector machines, neural networks, or survival forests to postoperative outcome prediction.
Ten studies applying ML models to esophagectomy outcomes were included (median cohort size 700; range 200–4700). Anastomotic leak was the most frequently predicted outcome (4/10), followed by mortality, major complications, readmission, strictures, and recurrence or survival. Common algorithms included gradient boosting (XGBoost, LightGBM, GBM), support vector machines, neural networks, and survival forests. Reported discrimination ranged from moderate to high (AUROC 0.64 for 90-day mortality and 0.65–0.70 for major complications, increasing to 0.79–0.90 for anastomotic leak prediction; some internally validated models reported AUROC >0.95). Three studies performed independent external validation, and performance generally declined in external cohorts. Comparative analyses demonstrated that ML often matched but did not consistently outperform regression-based models.
Machine learning models for postoperative risk prediction after esophagectomy demonstrate promising discrimination, particularly for anastomotic leak. Although external validation remains limited, ML approaches are still in early stages of clinical translation. With prospective data integration and robust multicenter validation, ML has the potential to enhance individualized risk stratification, support shared decision-making, guide perioperative planning, and improve allocation of postoperative resources in esophageal surgery.
T. Wang, Otari Beldishevski-Shotadze, N. Evennett· Diseases of the esophagus· 0 citations
Background Early identification of patients at high risk of anastomotic leak (AL) following esophagectomy is essential for improving surgical outcomes. However, reliable preoperative risk stratification remains challenging. This study aimed to predict AL risk in the esophageal cancer (EC) population by developing and validating a machine learning (ML)-based model using exclusively preoperative and baseline clinical data. Methods A retrospective cohort of EC patients who underwent radical esophagectomy at the Affiliated Tumor Hospital of Xinjiang Medical University from January 2020 to May 2025 was analyzed. Feature selection was performed using the least absolute shrinkage and selection operator (LASSO) regression. Five ML algorithms, Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM), were constructed. Model performance was comprehensively evaluated using the area under the receiver operating characteristic (ROC) curve (AUC), calibration curves, and decision curve analysis (DCA). Model interpretability was enhanced via Shapley Additive Explanations (SHAP), enabling quantification of feature importance and visualization of individual prediction contributions. Results A total of 368 patients were included and randomized to training (n=258) and validation (n=110) cohorts. Among the models, RF displayed the best discriminative performance in the validation cohort (AUC: 0.803, 95% confidence interval [CI]: 0.717-0.892), followed by XGBoost (AUC: 0.723) and LightGBM (AUC: 0.713). The RF model yielded a sensitivity (SEN) of 0.879 and a specificity (SPE) of 0.571. SHAP analysis identified monocytes, carcinoembryonic antigen (CEA), neutrophil-to-lymphocyte ratio (NLR), urine creatinine (UCr), and T stage as the five most influential predictors of AL. Calibration curves for the ensemble models demonstrated good agreement between predicted probabilities and observed outcomes. Conclusions The RF model, incorporating five routinely available preoperative variables, exhibited robust discriminative performance with high SEN for predicting in-hospital AL following esophagectomy. The proposed threshold-based risk stratification approach may facilitate individualized perioperative monitoring and management.
Yueying Yang, Kayishaer Ainiwaer, Yunfei Gao et al.· Frontiers in Oncology· 0 citations
Breast cancer recurrence remains a major cause of mortality among women worldwide. Early identification is essential for improving patient outcomes, and prediction modelling has emerged as an approach to support this objective. This study aimed to develop and compare machine learning models for predicting breast cancer recurrence. A retrospective secondary analysis was conducted on data from 286 patients in the University of California, Irvine (UCI) Machine Learning Repository. Nine predictors were used to develop logistic regression (LR), artificial neural network (ANN), and extreme gradient boosting (XGBoost) models. The dataset was split into training and testing sets using a 70:30 ratio. Model performance was evaluated using the area under the receiver operating characteristic curve (ROC-AUC), Brier score, and calibration analysis, with 95% confidence intervals estimated through bootstrap resampling. LR demonstrated the best discriminatory performance, achieving a test AUC of 0.790. In contrast, XGBoost and ANN showed lower generalization performance, with test AUCs of 0.727 and 0.748, respectively. LR also achieved the highest testing recall (0.731) and F1-score (0.623), indicating superior sensitivity for identifying recurrent cases. Meanwhile, XGBoost demonstrated the highest precision (0.727), accuracy (0.756), and calibration performance. The relatively small sample size and reliance on structured clinical predictors may have contributed to the superior performance of LR in this study. In conclusion, LR demonstrated the most reliable predictive performance for this dataset. Future research should use larger, diverse datasets and incorporate a broader range of predictors, including imaging and genomic data, to optimize the benefits of more complex machine learning models.
Nabila Shafiya, Soenarnatalina Melaniani, Ronny Isnuwardana et al.· Jurnal Biometrika dan Kepend...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
