BMC medical imaging · 2026
Predicting local control of brain metastases after Gamma Knife radiosurgery: a multimodal deep learning-radiomics approach.
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.
Zhao Minghui, Ding Xiaofeng, Sun Xinchen, Chen Qun, Li Jindong
Аннотация
BACKGROUND: Accurate prediction of local control (LC) following Gamma Knife radiosurgery (GKRS) remains clinically challenging for patients with brain metastases (BMs) secondary to breast or lung cancer. The scarcity of validated pre-therapeutic biomarkers currently hinders the development of individualized treatment decision-making strategies.
METHODS: This retrospective analysis included 324 BMs from 195 patients. To develop and benchmark various predictive frameworks, several models were constructed: clinical-only, radiomics-only, 2D deep learning (DL)-only, fused radiomics-DL (DLR), and a Combined model integrating all three modalities (clinical, radiomic, and DL features). All imaging signatures were derived from pre-treatment contrast-enhanced T1-weighted (T1CE) magnetic resonance imaging (MRI). Model performance was evaluated using the area under the curve (AUC), calibration plots, the Hosmer-Lemeshow test, and decision curve analysis (DCA).
RESULTS: The Combined model demonstrated consistently high numerical discriminative performance, achieving an AUC of 0.854 (95% Confidence Interval [CI]: 0.791-0.917) in the training cohort and 0.781 (95% CI: 0.618-0.945) in the independent test cohort. DCA demonstrated that the Combined model yielded superior net clinical benefit across threshold probabilities ranging from 10% to 50%, outperforming all other candidate models. Consequently, a nomogram based on the Combined model was developed to facilitate individualized risk estimation.
CONCLUSIONS: The integration of clinical parameters, handcrafted radiomic features, and 2D deep learning signatures provides a robust predictive tool for LC following GKRS in patients with brain metastases from breast or lung cancer. This multimodal approach holds significant potential for refining post-radiosurgical management and advancing personalized neuro-oncology care.
METHODS: This retrospective analysis included 324 BMs from 195 patients. To develop and benchmark various predictive frameworks, several models were constructed: clinical-only, radiomics-only, 2D deep learning (DL)-only, fused radiomics-DL (DLR), and a Combined model integrating all three modalities (clinical, radiomic, and DL features). All imaging signatures were derived from pre-treatment contrast-enhanced T1-weighted (T1CE) magnetic resonance imaging (MRI). Model performance was evaluated using the area under the curve (AUC), calibration plots, the Hosmer-Lemeshow test, and decision curve analysis (DCA).
RESULTS: The Combined model demonstrated consistently high numerical discriminative performance, achieving an AUC of 0.854 (95% Confidence Interval [CI]: 0.791-0.917) in the training cohort and 0.781 (95% CI: 0.618-0.945) in the independent test cohort. DCA demonstrated that the Combined model yielded superior net clinical benefit across threshold probabilities ranging from 10% to 50%, outperforming all other candidate models. Consequently, a nomogram based on the Combined model was developed to facilitate individualized risk estimation.
CONCLUSIONS: The integration of clinical parameters, handcrafted radiomic features, and 2D deep learning signatures provides a robust predictive tool for LC following GKRS in patients with brain metastases from breast or lung cancer. This multimodal approach holds significant potential for refining post-radiosurgical management and advancing personalized neuro-oncology care.