Applied radiation and isotopes : including data, instrumentation and methods for use in agriculture, industry and medicine · 2026
A novel automated segmentation strategy for clinical target volume of nasopharyngeal carcinoma based on multimodality and clinical processes simulation.
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 Zhuoxin, Yang Long, Zhong Yang, Shi Ziqi, Wang Jiazhou, Hu Chaosu, Hu Weigang
Аннотация
PURPOSE: Current automated segmentation methods for nasopharyngeal carcinoma clinical target volume (CTV) mainly rely on CT images and do not utilize multimodal imaging to simulate the process of clinical delineation of the target volume. The aim of this study was to enhance CTV auto-segmentation by utilizing information from multimodal imaging and simulating the clinical delineation process.
METHODS: To simulate the clinical delineation process of nasopharyngeal carcinoma CTV, we initially performed automatic segmentation of the gross tumor volume (GTV) in computed tomography (CT) and magnetic resonance imaging (MRI). Subsequently, this region of interest (ROI) was transferred to the CT image, where CTV segmentation was performed automatically based on the GTV contour and CT images. We employed the Swin UNETR network for automatic segmentation in both tasks. The model's performance was assessed using Dice Similarity Coefficient (DSC), Jaccard Index (JI), and 95% Hausdorff Distance (HD95) metrics and average symmetric surface distance (ASD). Two additional models were trained for comparison: CT-Only and CT-MRI fused segmentation. A five-fold cross-validation was conducted for performance evaluation.
RESULTS: The proposed method (CT + GTV) achieved superior segmentation performance, yielding a mean DSC of 0.844 (95% CI: 0.835-0.853), JI of 0.727 (95% CI: 0.715-0.739), HD95 of 7.96 mm (95% CI: 6.61-9.31), and ASD of 1.66 mm (95% CI: 1.48-1.85). This represented a significant improvement over the CT-Only baseline (DSC: 0.829; 95% CI: 0.818-0.840; HD95: 8.60 mm). Notably, the unstructured integration of MRI (CT + MRI) resulted in inferior performance (DSC: 0.818; 95% CI: 0.808-0.828; HD95: 9.89 mm) compared to using CT alone.
CONCLUSIONS: Simulating the clinical workflow by explicitly incorporating GTV spatial priors significantly enhances CTV auto-segmentation accuracy in NPC. This two-step strategy effectively leverages multimodal imaging, offering a more robust and clinically relevant alternative to direct image fusion for radiotherapy treatment planning.
METHODS: To simulate the clinical delineation process of nasopharyngeal carcinoma CTV, we initially performed automatic segmentation of the gross tumor volume (GTV) in computed tomography (CT) and magnetic resonance imaging (MRI). Subsequently, this region of interest (ROI) was transferred to the CT image, where CTV segmentation was performed automatically based on the GTV contour and CT images. We employed the Swin UNETR network for automatic segmentation in both tasks. The model's performance was assessed using Dice Similarity Coefficient (DSC), Jaccard Index (JI), and 95% Hausdorff Distance (HD95) metrics and average symmetric surface distance (ASD). Two additional models were trained for comparison: CT-Only and CT-MRI fused segmentation. A five-fold cross-validation was conducted for performance evaluation.
RESULTS: The proposed method (CT + GTV) achieved superior segmentation performance, yielding a mean DSC of 0.844 (95% CI: 0.835-0.853), JI of 0.727 (95% CI: 0.715-0.739), HD95 of 7.96 mm (95% CI: 6.61-9.31), and ASD of 1.66 mm (95% CI: 1.48-1.85). This represented a significant improvement over the CT-Only baseline (DSC: 0.829; 95% CI: 0.818-0.840; HD95: 8.60 mm). Notably, the unstructured integration of MRI (CT + MRI) resulted in inferior performance (DSC: 0.818; 95% CI: 0.808-0.828; HD95: 9.89 mm) compared to using CT alone.
CONCLUSIONS: Simulating the clinical workflow by explicitly incorporating GTV spatial priors significantly enhances CTV auto-segmentation accuracy in NPC. This two-step strategy effectively leverages multimodal imaging, offering a more robust and clinically relevant alternative to direct image fusion for radiotherapy treatment planning.