European journal of radiology · 2026
A biologically Interpretable CT radiomic model Predicts tertiary lymphoid structure status and Stratifies outcomes following chemoimmunotherapy in pancreatic ductal adenocarcinoma.
СтатьяНазвание и аннотация: Перевод в очереди
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.
Zhao Cancan, Zeng Chuyi, Huang Jingcheng, Shen Junjian, Zhou Jianjun, Li Dalong, Fang Shiwu, Zeng Mengsu, Zhu Qiuyang, Tang Qiying, Hong Minping, Xie Zongyu, Sun Haitao
Аннотация
PURPOSE: To develop and validate a CT-based radiomic model for non-invasive prediction of tertiary lymphoid structure (TLS) status in pancreatic ductal adenocarcinoma (PDAC), and to evaluate its biological correlates and association with clinical outcomes among patients receiving immune checkpoint inhibitors (ICIs) plus chemotherapy.
MATERIALS AND METHODS: A total of 435 patients with pathologically confirmed PDAC were retrospectively included in training (n = 170), temporal validation (n = 71), external validation (n = 63), radiogenomic validation (n = 50), and an immunotherapy cohort (n = 81). Radiomic features from preoperative CT were used to construct the RaTLS model with XGBoost after sequential feature selection. Radiogenomic analysis intersected differentially expressed genes between imaging- and pathology-defined TLS groups, and the resulting signature was projected into TCGA-PAAD for validation. Clinical outcomes were evaluated using objective response rate and progression-free survival.
RESULTS: The RaTLS model achieved AUCs of 0.891, 0.835, 0.794, and 0.722 across the four cohorts, with satisfactory calibration and positive decision-curve net benefit. Radiogenomic integration identified 522 concordant genes enriched in B cell and chemokine programs, and the corresponding signature was independently associated with overall survival in TCGA-PAAD (P = 0.002). RaTLS-high patients showed a higher objective response rate and longer progression-free survival, with RaTLS remaining independently associated with progression-free survival after multivariable adjustment (P = 0.029).
CONCLUSIONS: The CT-based RaTLS model shows promise for non-invasive prediction of TLS status in PDAC and may offer additional value in outcome stratification among patients receiving immunotherapy-based treatment.
MATERIALS AND METHODS: A total of 435 patients with pathologically confirmed PDAC were retrospectively included in training (n = 170), temporal validation (n = 71), external validation (n = 63), radiogenomic validation (n = 50), and an immunotherapy cohort (n = 81). Radiomic features from preoperative CT were used to construct the RaTLS model with XGBoost after sequential feature selection. Radiogenomic analysis intersected differentially expressed genes between imaging- and pathology-defined TLS groups, and the resulting signature was projected into TCGA-PAAD for validation. Clinical outcomes were evaluated using objective response rate and progression-free survival.
RESULTS: The RaTLS model achieved AUCs of 0.891, 0.835, 0.794, and 0.722 across the four cohorts, with satisfactory calibration and positive decision-curve net benefit. Radiogenomic integration identified 522 concordant genes enriched in B cell and chemokine programs, and the corresponding signature was independently associated with overall survival in TCGA-PAAD (P = 0.002). RaTLS-high patients showed a higher objective response rate and longer progression-free survival, with RaTLS remaining independently associated with progression-free survival after multivariable adjustment (P = 0.029).
CONCLUSIONS: The CT-based RaTLS model shows promise for non-invasive prediction of TLS status in PDAC and may offer additional value in outcome stratification among patients receiving immunotherapy-based treatment.