European journal of radiology open · 2026
DCE-MRI and Mammography integrated radiomic analysis in triple negative ductal invasive breast cancer patients. Comparison between BRCA and not BRCA mutated patients: preliminary results.
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.
Pecchi Annarita, Beretta Cecilia, Balboni Erica, Nocetti Luca, Bozzola Chiara, Sessa Giulia, Toss Angela, Cortesi Laura, Guidi Gabriele, Dominici Massimo, Torricelli Pietro
Аннотация
PURPOSE: We propose a combined radiomic model based on features extracted from DCE-MRI and mammographic images (MG) to discriminate the mutational status of BRCA genes in patients with TNBC.
MATERIAL AND METHODS: This retrospective study included patients histologically diagnosed with TNBC who performed a mammography followed by a DCE-MRI between 2010 and 2021.A manual segmentation was performed on the entire tumor volume on both imaging modalities. Subsequently, a 5-mm ROI was placed within the segmented tumor volume and the contralateral healthy gland. Radiomic features were extracted from these ROIs using Pyradiomics and selected with Maximum Relevance Minimum Redundancy algorithm to fit different classifiers.A final model was developed by integrating the most predictive features extracted from lesion and healthy gland from both modalities into a single multivariate framework. Finally, the added value of an integrated approach was assessed by comparing the performance with single-modality radiomic models.
RESULTS: The population included 52 patients for a total of 53 lesions (13 BRCA-mutated, 40 non BRCA-carriers). The highest classification performance for BRCA mutational status was achieved by Random Forest that integrated both MRI and mammography features, yielding an Area Under the Curve (AUC) of 0.91. Most predictive features were extracted from healthy gland regions and from MR wash-in/wash-out maps. Multimodal imaging analysis outperformed the best models of single-modality approaches (AUCMG-gland 0.827, AUCMR-gland+lesion 0.80).
CONCLUSIONS: This study reinforces the feasibility of radiomics to predict BRCA mutation in patients with TNBC, highlighting the possibility of integrating different imaging modalities to enhance model accuracy.
MATERIAL AND METHODS: This retrospective study included patients histologically diagnosed with TNBC who performed a mammography followed by a DCE-MRI between 2010 and 2021.A manual segmentation was performed on the entire tumor volume on both imaging modalities. Subsequently, a 5-mm ROI was placed within the segmented tumor volume and the contralateral healthy gland. Radiomic features were extracted from these ROIs using Pyradiomics and selected with Maximum Relevance Minimum Redundancy algorithm to fit different classifiers.A final model was developed by integrating the most predictive features extracted from lesion and healthy gland from both modalities into a single multivariate framework. Finally, the added value of an integrated approach was assessed by comparing the performance with single-modality radiomic models.
RESULTS: The population included 52 patients for a total of 53 lesions (13 BRCA-mutated, 40 non BRCA-carriers). The highest classification performance for BRCA mutational status was achieved by Random Forest that integrated both MRI and mammography features, yielding an Area Under the Curve (AUC) of 0.91. Most predictive features were extracted from healthy gland regions and from MR wash-in/wash-out maps. Multimodal imaging analysis outperformed the best models of single-modality approaches (AUCMG-gland 0.827, AUCMR-gland+lesion 0.80).
CONCLUSIONS: This study reinforces the feasibility of radiomics to predict BRCA mutation in patients with TNBC, highlighting the possibility of integrating different imaging modalities to enhance model accuracy.