Predicting liver metastasis in pancreatic neuroendocrine tumors with an interpretable machine learning algorithm: a SEER-based study DOI Creative Commons

Jinzhe Bi,

Yaqun Yu

Frontiers in Medicine, Journal Year: 2025, Volume and Issue: 12

Published: May 1, 2025

Liver metastasis is the most common site of in pancreatic neuroendocrine tumors (PaNETs), significantly affecting patient prognosis. This study aims to develop machine learning algorithms predict liver PaNETs patients, assisting clinicians personalized clinical decision-making for treatment. We collected data on eligible patients from Surveillance, Epidemiology, and End Results (SEER) database period 2010 2021. The Boruta algorithm Least Absolute Shrinkage Selection Operator (LASSO) were used feature selection. applied 10 different models predicting risk patients. model's performance was assessed using a variety metrics, including area under receiver operating characteristic curve (AUC), precision-recall (AUPRC), decision analysis (DCA), calibration curves, accuracy, sensitivity, specificity, F1 score, Kappa score. SHapley Additive exPlanations (SHAP) employed interpret models, best-performing model web-based calculator. included cohort 7,463 whom 1,356 (18.2%) diagnosed with at time initial diagnosis. Through combined use LASSO methods, T-stage, N-stage, tumor size, grade, surgery, lymphadenectomy, chemotherapy, bone identified as independent factors PaNETs. Compared other algorithms, gradient boosting (GBM) exhibited superior performance, achieving an AUC 0.937 (95% CI: 0.931-0.943), AUPRC 0.94, accuracy 0.87. DCA analyses demonstrate that GBM provides better capabilities predictive performance. Furthermore, SHAP framework revealed T-stage are primary influencing predictions. Finally, based algorithm, we developed accessible calculator excels outperforming providing critical support developing medical strategies practice.

Language: Английский

Predicting liver metastasis in pancreatic neuroendocrine tumors with an interpretable machine learning algorithm: a SEER-based study DOI Creative Commons

Jinzhe Bi,

Yaqun Yu

Frontiers in Medicine, Journal Year: 2025, Volume and Issue: 12

Published: May 1, 2025

Liver metastasis is the most common site of in pancreatic neuroendocrine tumors (PaNETs), significantly affecting patient prognosis. This study aims to develop machine learning algorithms predict liver PaNETs patients, assisting clinicians personalized clinical decision-making for treatment. We collected data on eligible patients from Surveillance, Epidemiology, and End Results (SEER) database period 2010 2021. The Boruta algorithm Least Absolute Shrinkage Selection Operator (LASSO) were used feature selection. applied 10 different models predicting risk patients. model's performance was assessed using a variety metrics, including area under receiver operating characteristic curve (AUC), precision-recall (AUPRC), decision analysis (DCA), calibration curves, accuracy, sensitivity, specificity, F1 score, Kappa score. SHapley Additive exPlanations (SHAP) employed interpret models, best-performing model web-based calculator. included cohort 7,463 whom 1,356 (18.2%) diagnosed with at time initial diagnosis. Through combined use LASSO methods, T-stage, N-stage, tumor size, grade, surgery, lymphadenectomy, chemotherapy, bone identified as independent factors PaNETs. Compared other algorithms, gradient boosting (GBM) exhibited superior performance, achieving an AUC 0.937 (95% CI: 0.931-0.943), AUPRC 0.94, accuracy 0.87. DCA analyses demonstrate that GBM provides better capabilities predictive performance. Furthermore, SHAP framework revealed T-stage are primary influencing predictions. Finally, based algorithm, we developed accessible calculator excels outperforming providing critical support developing medical strategies practice.

Language: Английский

Citations

0