Journal of advanced research · 2026
Multi-modal gradual fusion transformer-based model for predicting immunotherapy response in patients with hepatocellular carcinoma.
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.
Xiao Lushan, Wang Jiaren, Cui Hao, Zhu Hongbo, He Jingzhe, Deng Huangying, Zhang Wenfeng, Dong Hanzhi, Zhou Yundong, Jiang Pu, Zeng Lin, Peng Jie, Xu Peishuang, Shen Ruozheng, Kurban Nazakat, Lin Manxun, Lu Shaodong, Weng Xie, Hong Chang, Liu Li
Аннотация
BACKGROUND: Immunotherapy effectively extends survival in hepatocellular carcinoma (HCC) patients. Predicting immunotherapy responses can inform treatment strategies for HCC. This study aimed to develop multi-modal transformer-based models to predict the immunotherapy response and to validate their performance in an independent cohort.
MATERIALS AND METHODS: Patients with HCC from five medical centers were retrospectively included. Clinical features were selected using Least Absolute Shrinkage and Selection Operator method. Multi-modal gradual fusion transformer-based models were trained using clinical features and intra- and peritumoral patches from arterial and portal venous phase computed tomography images in the training cohort. These models were tested on internal validation and external test cohorts. Models' performance and generalization across different modalities were compared.
RESULTS: Patients from the Hospital 1 were partitioned into a training (n = 209) and an internal validation cohort (n = 90) at a 7:3 ratio. And patients from the other four centers formed an independent external test cohort (n = 85). The number of progressive disease (PD) patients in the training, internal validation, and external test cohorts was 44 (21.1%), 20 (22.2%), and 20 (23.5%), respectively. The model using clinical data, intratumoral imaging, and peritumoral imaging modalities (GIFT-CIP) demonstrated strong predictive performance, achieving an area under the curve (AUC) values of 0.926 (95% CI: 0.892-0.962), 0.911 (95% CI: 0.878-0.946), and 0.883 (95% CI: 0.835-0.935) for training cohort, internal validation cohort, and external test cohort, respectively. Crucially, the GIFT-CIP model effectively stratified patients into low- and high-risk groups, showing significant differences in progression-free survival and overall survival in external test cohort (p < 0.01).
CONCLUSIONS: The GIFT-CIP model is a non-invasive method for predicting immunotherapy responses in patients with HCC. This model may be clinically useful for assisting clinicians in guiding surveillance follow-up and identifying optimal immunotherapy strategies.
MATERIALS AND METHODS: Patients with HCC from five medical centers were retrospectively included. Clinical features were selected using Least Absolute Shrinkage and Selection Operator method. Multi-modal gradual fusion transformer-based models were trained using clinical features and intra- and peritumoral patches from arterial and portal venous phase computed tomography images in the training cohort. These models were tested on internal validation and external test cohorts. Models' performance and generalization across different modalities were compared.
RESULTS: Patients from the Hospital 1 were partitioned into a training (n = 209) and an internal validation cohort (n = 90) at a 7:3 ratio. And patients from the other four centers formed an independent external test cohort (n = 85). The number of progressive disease (PD) patients in the training, internal validation, and external test cohorts was 44 (21.1%), 20 (22.2%), and 20 (23.5%), respectively. The model using clinical data, intratumoral imaging, and peritumoral imaging modalities (GIFT-CIP) demonstrated strong predictive performance, achieving an area under the curve (AUC) values of 0.926 (95% CI: 0.892-0.962), 0.911 (95% CI: 0.878-0.946), and 0.883 (95% CI: 0.835-0.935) for training cohort, internal validation cohort, and external test cohort, respectively. Crucially, the GIFT-CIP model effectively stratified patients into low- and high-risk groups, showing significant differences in progression-free survival and overall survival in external test cohort (p < 0.01).
CONCLUSIONS: The GIFT-CIP model is a non-invasive method for predicting immunotherapy responses in patients with HCC. This model may be clinically useful for assisting clinicians in guiding surveillance follow-up and identifying optimal immunotherapy strategies.