Quantitative imaging in medicine and surgery · 2026
Deep learning-assisted prediction of hydrocephalus in preoperative-subarachnoid hemorrhage: a multi-center study.
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.
Han Wenrui, Yang Cheng, Liu Chengli, Wang Guijun, Tian Qi, Zhang Zhongyang, Liao Jianming, Li Mingchang
Аннотация
BACKGROUND: Delayed hydrocephalus, a complication that significantly impacts patient prognosis, arises following subarachnoid hemorrhage (SAH). In this study, we conducted a comprehensive investigation into the correlation between clinical features and computed tomography (CT) images, with the aim of elucidating the development of delayed hydrocephalus in SAH patients. We have developed a model for the early detection and evaluation of SAH, which is based on deep learning (DL).
METHODS: The DeepSH model was constructed by concatenating a super-resolution generative adversarial network (SRGAN) model based on a convolutional neural network (CNN) generator and an imagingomics model based on traditional machine learning (ML) models. The data were obtained from non-contrast CT (NCCT) images of 861 patients with SAH admitted to three hospitals in China between July 2019 and December 2021. After training, the model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curves. The ROC curves and calibration curves indicated that the model fit well, and the decision curve analysis (DCA) validated the clinical utility of DeepSH.
RESULTS: DeepSH demonstrated superior performance, with areas under the curves (AUCs) improved by 0.006 and 0.219 (P<0.05) in the internal- and external-testing sets, respectively, relative to the support vector machine (SVM) model. Clinical benefit and overall efficiency of junior radiologists were significantly improved with model assistance for the internal- and external-testing sets.
CONCLUSIONS: DeepSH significantly outperforms conventional methods and expert assessment in predicting SAH-associated delayed hydrocephalus (SAH-H) from NCCT images, providing a valuable tool for early clinical decision-making that can improve patient prognosis.
METHODS: The DeepSH model was constructed by concatenating a super-resolution generative adversarial network (SRGAN) model based on a convolutional neural network (CNN) generator and an imagingomics model based on traditional machine learning (ML) models. The data were obtained from non-contrast CT (NCCT) images of 861 patients with SAH admitted to three hospitals in China between July 2019 and December 2021. After training, the model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curves. The ROC curves and calibration curves indicated that the model fit well, and the decision curve analysis (DCA) validated the clinical utility of DeepSH.
RESULTS: DeepSH demonstrated superior performance, with areas under the curves (AUCs) improved by 0.006 and 0.219 (P<0.05) in the internal- and external-testing sets, respectively, relative to the support vector machine (SVM) model. Clinical benefit and overall efficiency of junior radiologists were significantly improved with model assistance for the internal- and external-testing sets.
CONCLUSIONS: DeepSH significantly outperforms conventional methods and expert assessment in predicting SAH-associated delayed hydrocephalus (SAH-H) from NCCT images, providing a valuable tool for early clinical decision-making that can improve patient prognosis.