International journal of surgery (London, England) · 2025
Explainable multiparameter MRI radiomics model for the noninvasive tracing of the origin of vertebral metastatic cancer: a multicenter cohort study.
СтатьяМногоцентровое исследованиеНазвание 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.
Zhang Hao, Yuan Genji, Wu Ruochen, Zhang Ziyue, Xu Ruixiang, Xu Tongshuai, Zhong Xin, Kong Meng, Wang Yan, Tao Hao, Ma Xuexiao
Аннотация
BACKGROUND: The accurate identification of the primary origin of malignant vertebral compression fractures and vertebral metastatic cancer (VMC) is critical for effective treatment, but current diagnostic approaches are invasive and lack specificity. We developed an explainable MRI-based radiomics model for tracing VMC origins noninvasively.
MATERIALS AND METHODS: This multicenter study included 1123 patients with vertebral compression fractures from 5 hospitals with confirmed diagnoses; data from 754 patients in 3 centers were divided into training and validation cohorts based on the date of MRI examination. Two independent external test cohorts were used to evaluate the generalizability of the model. Sagittal T1-weighted images, T2 weighted images, and fat-suppression T2-weighted images were used for manual 3D lesion segmentation. In total, 3135 radiomic features were extracted from each region of interest. Feature selection was performed using the minimal redundancy maximal relevance algorithm. The final prediction model was established using a support vector machine algorithm.
RESULTS: The radiomics model achieved areas under the receiver operating characteristic curves of 0.99, 0.87, and 0.90 in the validation, and two external test datasets, respectively, for VMC detection. For primary origin prediction, the model demonstrated high diagnostic accuracy across all datasets for lung, breast, and prostate origins. DeLong's test indicated a strong generalization capability of the model. SHAP analysis identified wavelet as key contributors to model transparency.
CONCLUSION: Our model was capable of distinguishing VMC from benign lesions and accurately predicting its primary origin. It demonstrated high generalizability and interpretability and represents a promising noninvasive tool to support clinical decision making and reduce invasive biopsies.
MATERIALS AND METHODS: This multicenter study included 1123 patients with vertebral compression fractures from 5 hospitals with confirmed diagnoses; data from 754 patients in 3 centers were divided into training and validation cohorts based on the date of MRI examination. Two independent external test cohorts were used to evaluate the generalizability of the model. Sagittal T1-weighted images, T2 weighted images, and fat-suppression T2-weighted images were used for manual 3D lesion segmentation. In total, 3135 radiomic features were extracted from each region of interest. Feature selection was performed using the minimal redundancy maximal relevance algorithm. The final prediction model was established using a support vector machine algorithm.
RESULTS: The radiomics model achieved areas under the receiver operating characteristic curves of 0.99, 0.87, and 0.90 in the validation, and two external test datasets, respectively, for VMC detection. For primary origin prediction, the model demonstrated high diagnostic accuracy across all datasets for lung, breast, and prostate origins. DeLong's test indicated a strong generalization capability of the model. SHAP analysis identified wavelet as key contributors to model transparency.
CONCLUSION: Our model was capable of distinguishing VMC from benign lesions and accurately predicting its primary origin. It demonstrated high generalizability and interpretability and represents a promising noninvasive tool to support clinical decision making and reduce invasive biopsies.