Journal of the National Cancer Center · 2026
Construction and external validation of radiomics models to detect primary prostate cancer with machine learning: a multicenter study based on 68Ga-PSMA PET/CT.
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 Jiaxian, Jiang Xin, Yang Guangjie, Tang Yongxiang, Qi Lin, Chen Minfeng, Hu Shuo, Gao Xiaomei, Gan Yu, Zhang Mingxin, Chen Shouzhen, Cai Yi
Аннотация
OBJECTIVE: Recent studies have shown that 68Ga-prostate-specific membrane antigen (68Ga-PSMA) positron emission tomography/computed tomography (PET/CT) can open a non-invasive diagnostic pathway in prostate cancer (PCa). However, the relatively small number of enrolled patients in these studies limits statistical validation and causes bias to some extent. In addition, the performance of 68Ga-PSMA PET/CT radiomics analysis in detecting PCa has not been widely evaluated. Hence, the present multicenter study endeavored to develop and validate radiomics models based on 68Ga-PSMA PET/CT for the detection of primary PCa in a relatively larger cohort.
METHODS: This study enrolled consecutive patients with suspected PCa who underwent systematic biopsy (SB) and PSMA PET/CT-targeted biopsy (TB) after 68Ga-PSMA PET/CT in three medical centers. The whole prostate gland was adopted as the volume of interest (VOI). Eight machine learning (ML) algorithms were utilized to develop models with selected radiomics features separately, after which the best-performing radiomics model was integrated with maximum standardized uptake value (SUVmax) to create the combined model. The receiver operating characteristic (ROC) curves and area under the curve (AUC) value, accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for each model.
RESULTS: Overall, 609 patients were recruited, including 195 patients with benign prostate diseases (BPD), 30 patients with clinically insignificant prostate cancer (ciPCa, Gleason score [GS] = 3 + 3), and 384 patients with clinically significant prostate cancer (csPCa, GS ≥ 3 + 4). For csPCa prediction, the radiomics model developed by the eXtreme Gradient Boosting (XGBoost) algorithm demonstrated the best performance. After integrating with SUVmax, the combined model achieved the highest AUC of 0.921 in internal validation cohort. External validation cohort 1 and 2 also showed promising results with AUC of 0.906 and 0.898, respectively. For PCa prediction, the XGBoost algorithm combined with SUVmax also peformed well in three validation cohorts with the AUC ranging from 0.860 to 0.918.
CONCLUSIONS: This is the largest multicenter study to date, providing a noninvasive and quantitative method based on 68Ga-PSMA PET/CT radiomics analysis modeled with ML for predicting PCa. This method has the potential to gauge the risk of PCa before biopsy.
METHODS: This study enrolled consecutive patients with suspected PCa who underwent systematic biopsy (SB) and PSMA PET/CT-targeted biopsy (TB) after 68Ga-PSMA PET/CT in three medical centers. The whole prostate gland was adopted as the volume of interest (VOI). Eight machine learning (ML) algorithms were utilized to develop models with selected radiomics features separately, after which the best-performing radiomics model was integrated with maximum standardized uptake value (SUVmax) to create the combined model. The receiver operating characteristic (ROC) curves and area under the curve (AUC) value, accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for each model.
RESULTS: Overall, 609 patients were recruited, including 195 patients with benign prostate diseases (BPD), 30 patients with clinically insignificant prostate cancer (ciPCa, Gleason score [GS] = 3 + 3), and 384 patients with clinically significant prostate cancer (csPCa, GS ≥ 3 + 4). For csPCa prediction, the radiomics model developed by the eXtreme Gradient Boosting (XGBoost) algorithm demonstrated the best performance. After integrating with SUVmax, the combined model achieved the highest AUC of 0.921 in internal validation cohort. External validation cohort 1 and 2 also showed promising results with AUC of 0.906 and 0.898, respectively. For PCa prediction, the XGBoost algorithm combined with SUVmax also peformed well in three validation cohorts with the AUC ranging from 0.860 to 0.918.
CONCLUSIONS: This is the largest multicenter study to date, providing a noninvasive and quantitative method based on 68Ga-PSMA PET/CT radiomics analysis modeled with ML for predicting PCa. This method has the potential to gauge the risk of PCa before biopsy.