Revista espanola de medicina nuclear e imagen molecular · 2026
Preoperative prediction of lymph node metastasis and prognosis in gastric cancer patients: A machine learning model based on 18F-FDG metabolic parameters.
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.
Chen D, Chen W, Pan X, Lyu Q, Zhang J, Lin D, Liu D
Аннотация
OBJECTIVE: To construct and validate a model for predicting lymph node metastasis (LNMs) of gastric cancer (GC) based on 18F-FDG PET/CT multi-parameter data and multiple machine learning (ML) algorithms.
METHODS: We conducted a retrospective analysis of clinical records and 18F-FDG PET/CT images from 167 GC patients treated with radical gastrectomy. The dataset was randomly divided into two cohorts (training and validation cohort) with a ratio of 7:3. Seven machine learning algorithms were used to develop models. The model's performance was evaluated using metrics such as the receiver operating characteristic curve (ROC), sensitivity, specificity, accuracy, negative and positive predictive value, F1 score, calibration curve, and decision curve. Shapley Additive Explanations (SHAP) analysis was conducted to determine feature importance and enhance model interpretability. Additionally, Kaplan-Meier analysis was used to evaluate the prognosis of GC patients.
RESULTS: The Artificial Neural Network algorithm(ANN)model outperformed other algorithms in the validation set, with an AUC of 0.814 (95% CI: 0.707-0.921), a sensitivity of 0.872, a specificity of 0.500 and an accuracy of 0.784. Both calibration and decision curve analyses demonstrated excellent calibrated ability and favorable clinical applicability. Through SHAP method, tumor invasion depth (clinical T stage), platelet count (PLT), and total lesion glycolysis (TLG) emerged as the most influential predictors in the ML model. Furthermore, the model exhibited robust risk stratification capability for overall survival (OS).
CONCLUSION: The ML model utilizing 18F-FDG PET/CT metabolic parameters showed good efficacy and reliability in predicting LNS preoperatively in GC patients.
METHODS: We conducted a retrospective analysis of clinical records and 18F-FDG PET/CT images from 167 GC patients treated with radical gastrectomy. The dataset was randomly divided into two cohorts (training and validation cohort) with a ratio of 7:3. Seven machine learning algorithms were used to develop models. The model's performance was evaluated using metrics such as the receiver operating characteristic curve (ROC), sensitivity, specificity, accuracy, negative and positive predictive value, F1 score, calibration curve, and decision curve. Shapley Additive Explanations (SHAP) analysis was conducted to determine feature importance and enhance model interpretability. Additionally, Kaplan-Meier analysis was used to evaluate the prognosis of GC patients.
RESULTS: The Artificial Neural Network algorithm(ANN)model outperformed other algorithms in the validation set, with an AUC of 0.814 (95% CI: 0.707-0.921), a sensitivity of 0.872, a specificity of 0.500 and an accuracy of 0.784. Both calibration and decision curve analyses demonstrated excellent calibrated ability and favorable clinical applicability. Through SHAP method, tumor invasion depth (clinical T stage), platelet count (PLT), and total lesion glycolysis (TLG) emerged as the most influential predictors in the ML model. Furthermore, the model exhibited robust risk stratification capability for overall survival (OS).
CONCLUSION: The ML model utilizing 18F-FDG PET/CT metabolic parameters showed good efficacy and reliability in predicting LNS preoperatively in GC patients.