Frontiers in oncology · 2026
Explainable machine-learning model for postoperative lower-extremity deep vein thrombosis after gynecologic cancer surgery.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Zhang Yanhong, Shi Congcong, Yang Deyan, Ji Xiaona, Chu Yingli
Аннотация
BACKGROUND: Postoperative lower-extremity deep vein thrombosis (DVT) is a common and potentially fatal complication after radical surgery for gynecologic malignancies, yet population-specific risk tools remain limited.
METHODS: We retrospectively analyzed 500 consecutive patients who underwent radical surgery for cervical, endometrial, or ovarian cancer (January 2022-January 2025). The cohort was randomly divided into a training set (n=351, 70%) and a validation set (n=149, 30%) at a 7:3 ratio. Independent predictors were identified using logistic regression, and 10 machine-learning algorithms were trained and compared. Model discrimination, calibration, and clinical utility were evaluated, and the optimal model was interpreted using SHAP.
RESULTS: DVT occurred in 116 patients (23.2%). Five variables were independently associated with DVT: age, Caprini score, lymphocyte count, preoperative D-dimer, and neutrophil-to-lymphocyte ratio (NLR).Across candidate models, CatBoost achieved the optimal overall performance on the validation set, yielding an AUC of 0.816 (95% CI: 0.738-0.881), with sensitivity of 0.824 (95% CI: 0.700-1.000) and specificity of 0.730 (95% CI: 0.515-0.876), and SHAP analysis ranked the predictors by mean absolute SHAP value in the following order: lymphocyte count, Caprini score, age, preoperative D-dimer, and neutrophil-to-lymphocyte ratio (NLR).
CONCLUSION: An interpretable CatBoost-based model enables individualized postoperative DVT risk stratification after gynecologic cancer surgery and may support targeted, evidence-based thromboprophylaxis and perioperative nursing management.
METHODS: We retrospectively analyzed 500 consecutive patients who underwent radical surgery for cervical, endometrial, or ovarian cancer (January 2022-January 2025). The cohort was randomly divided into a training set (n=351, 70%) and a validation set (n=149, 30%) at a 7:3 ratio. Independent predictors were identified using logistic regression, and 10 machine-learning algorithms were trained and compared. Model discrimination, calibration, and clinical utility were evaluated, and the optimal model was interpreted using SHAP.
RESULTS: DVT occurred in 116 patients (23.2%). Five variables were independently associated with DVT: age, Caprini score, lymphocyte count, preoperative D-dimer, and neutrophil-to-lymphocyte ratio (NLR).Across candidate models, CatBoost achieved the optimal overall performance on the validation set, yielding an AUC of 0.816 (95% CI: 0.738-0.881), with sensitivity of 0.824 (95% CI: 0.700-1.000) and specificity of 0.730 (95% CI: 0.515-0.876), and SHAP analysis ranked the predictors by mean absolute SHAP value in the following order: lymphocyte count, Caprini score, age, preoperative D-dimer, and neutrophil-to-lymphocyte ratio (NLR).
CONCLUSION: An interpretable CatBoost-based model enables individualized postoperative DVT risk stratification after gynecologic cancer surgery and may support targeted, evidence-based thromboprophylaxis and perioperative nursing management.