European journal of radiology · 2026
Reproducibility of segmentation and radiomic features in contrast-enhanced CT images of early, intermediate and advanced HCC patients: a multicenter study.
СтатьяМногоцентровое исследованиеНазвание 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.
Fanni Salvatore Claudio, Maria Gabriele Di, Aghakhanyan Gayane, Braccischi Lorenzo, Celsa Ciro, Cannella Roberto, Cabibbo Giuseppe, Enea Marco, Francischello Roberto, Masi Gianluca, Matranga Domenica, Marseglia Mariarosaria, Mosconi Cristina, Pecorelli Anna, Salani Francesca, Stefanini Bernardo, Vivaldi Caterina, Cammà Calogero, Piscaglia Fabio, Lencioni Riccardo, Cioni Dania
Аннотация
BACKGROUND: Radiomics holds promise for extracting quantitative biomarkers from CT images of hepatocellular carcinoma (HCC), but its clinical translation is hampered by poor reproducibility caused by variations in image acquisition, manual segmentation and feature extraction. This multicenter study assessed the reproducibility of both segmentation and radiomic features in patients with early, intermediate and advanced HCC.
METHODS: Forty-five patients (15 per centre; 5 per tumor stage) from three institutions underwent contrast-enhanced CT. Three radiologists independently delineated lesion and perilesion regions on arterial (HAP) and portal venous phases (PVP) using 3D‑Slicer. Segmentation quality was assessed with the Jaccard index; masks with < 25% concordance were manually revised. Using the SlicerRadiomics module, 851 features per phase were extracted, and reproducibility was measured via the generalized concordance correlation coefficient (GCCC) before and after revision.
RESULTS: Median Jaccard values were 0,40, 0,56and 0,74, between center 1 and 2, between center 1 and center 3, and between center 2 and 3, respectively. Twenty-seven segmentations underwent manual revision based on Jaccard index. After manual revision Lesion features (mean GCCC 0,65) were more reproducible than perilesional ones (0,52), and PVP produced slightly higher concordance than HAP. Before revision, mean GCCC ranged from 0.45 to 0.68 (median 0.48-0.77); early-stage lesions had the highest reproducibility and advanced-stage the lowest. After revision, lesion mean GCCC improved to 0.71 (HAP) and 0.67 (PVP), with larger gains in intermediate-stage tumours, while perilesional concordance improved modestly.
CONCLUSIONS: Radiomic feature reproducibility depends on HCC stage and region. Early HCC lesions yield more reproducible features compared to advanced tumours. Manual revision enhances concordance, particularly for lesion regions, but defining peritumoral tissue remains challenging.
METHODS: Forty-five patients (15 per centre; 5 per tumor stage) from three institutions underwent contrast-enhanced CT. Three radiologists independently delineated lesion and perilesion regions on arterial (HAP) and portal venous phases (PVP) using 3D‑Slicer. Segmentation quality was assessed with the Jaccard index; masks with < 25% concordance were manually revised. Using the SlicerRadiomics module, 851 features per phase were extracted, and reproducibility was measured via the generalized concordance correlation coefficient (GCCC) before and after revision.
RESULTS: Median Jaccard values were 0,40, 0,56and 0,74, between center 1 and 2, between center 1 and center 3, and between center 2 and 3, respectively. Twenty-seven segmentations underwent manual revision based on Jaccard index. After manual revision Lesion features (mean GCCC 0,65) were more reproducible than perilesional ones (0,52), and PVP produced slightly higher concordance than HAP. Before revision, mean GCCC ranged from 0.45 to 0.68 (median 0.48-0.77); early-stage lesions had the highest reproducibility and advanced-stage the lowest. After revision, lesion mean GCCC improved to 0.71 (HAP) and 0.67 (PVP), with larger gains in intermediate-stage tumours, while perilesional concordance improved modestly.
CONCLUSIONS: Radiomic feature reproducibility depends on HCC stage and region. Early HCC lesions yield more reproducible features compared to advanced tumours. Manual revision enhances concordance, particularly for lesion regions, but defining peritumoral tissue remains challenging.