Nuclear medicine communications · 2025
Malignancy classification of thyroid incidentalomas using 18 F-fluorodeoxy- d -glucose PET/computed tomography-derived radiomics.
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.
Yeghaian Melda, Piek Marceline W, Bartels-Rutten Annemarieke, Abdelatty Mohamed A, Herrero-Huertas Marina, Vogel Wouter V, de Boer Jan Paul, Hartemink Koen J, Bodalal Zuhir, Beets-Tan Regina G H, Trebeschi Stefano, van der Ploeg Iris M C
Аннотация
BACKGROUND: Thyroid incidentalomas (TIs) are incidental thyroid lesions detected on fluorodeoxy- d -glucose ( 18 F-FDG) PET/computed tomography (PET/CT) scans. This study aims to investigate the role of noninvasive PET/CT-derived radiomic features in characterizing 18 F-FDG PET/CT TIs and distinguishing benign from malignant thyroid lesions in oncological patients.
MATERIALS AND METHODS: We included 46 patients with PET/CT TIs who underwent thyroid ultrasound and thyroid surgery at our oncological referral hospital. Radiomic features extracted from regions of interest (ROI) in both PET and CT images and analyzed for their association with thyroid cancer and their predictive ability. The TIs were graded using the ultrasound TIRADS classification, and histopathological results served as the reference standard. Univariate and multivariate analyses were performed using features from each modality individually and combined. The performance of radiomic features was compared to the TIRADS classification.
RESULTS: Among the 46 included patients, 36 patients (78%) had malignant thyroid lesions, while 10 patients (22%) had benign lesions. The combined run length nonuniformity radiomic feature from PET and CT cubical ROIs demonstrated the highest area under the curve (AUC) of 0.88 ( P < 0.05), with a negative correlation with malignancy. This performance was comparable to the TIRADS classification (AUC: 0.84, P < 0.05), which showed a positive correlation with thyroid cancer. Multivariate analysis showed higher predictive performance using CT-derived radiomics (AUC: 0.86 ± 0.13) compared to TIRADS (AUC: 0.80 ± 0.08).
CONCLUSION: This study highlights the potential of 18 F-FDG PET/CT-derived radiomics to distinguish benign from malignant thyroid lesions. Further studies with larger cohorts and deep learning-based methods could obtain more robust results.
MATERIALS AND METHODS: We included 46 patients with PET/CT TIs who underwent thyroid ultrasound and thyroid surgery at our oncological referral hospital. Radiomic features extracted from regions of interest (ROI) in both PET and CT images and analyzed for their association with thyroid cancer and their predictive ability. The TIs were graded using the ultrasound TIRADS classification, and histopathological results served as the reference standard. Univariate and multivariate analyses were performed using features from each modality individually and combined. The performance of radiomic features was compared to the TIRADS classification.
RESULTS: Among the 46 included patients, 36 patients (78%) had malignant thyroid lesions, while 10 patients (22%) had benign lesions. The combined run length nonuniformity radiomic feature from PET and CT cubical ROIs demonstrated the highest area under the curve (AUC) of 0.88 ( P < 0.05), with a negative correlation with malignancy. This performance was comparable to the TIRADS classification (AUC: 0.84, P < 0.05), which showed a positive correlation with thyroid cancer. Multivariate analysis showed higher predictive performance using CT-derived radiomics (AUC: 0.86 ± 0.13) compared to TIRADS (AUC: 0.80 ± 0.08).
CONCLUSION: This study highlights the potential of 18 F-FDG PET/CT-derived radiomics to distinguish benign from malignant thyroid lesions. Further studies with larger cohorts and deep learning-based methods could obtain more robust results.