International journal of surgery (London, England) · 2026
Deep learning-enhanced quantitative evaluation of hepatic ischemia-reperfusion injury and steatosis via near-infrared hyperspectral imaging.
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.
Wang Zichen, Qu Zhan, Chen Chen, Du Jian, Zhang Jian, Li Tao, Shi Jianhua, Zhang Chun, Lu Jianwen, Tian Min, Lu Lu, Tian Boyan, Zhao Zhen, Mu Fan, Xu Lexuan, Zhang Xin, Wang Yimeng, Yao Cuiping, Sun Hao, Lyu Yi, Wang Bo, Hu Liangshuo
Аннотация
INTRODUCTION: Global organ scarcity necessitates use of marginal liver grafts. We evaluated an independently developed noninvasive near-infrared hyperspectral imaging (NIR-HSI) system to assess hepatic steatosis, ischemia-reperfusion (I/R) injury, and inflammatory biomarkers in animal models and transplant recipients. Deep learning integration enhanced spectral discrimination, improving classification and regression accuracy.
MATERIAL AND METHODS: NIR-HSI data were used in conjunction with a one-dimensional convolutional neural network (1D-CNN) to classify the degree of hepatic steatosis and hepatic I/R injury. Additionally, least squares support vector machines (LSSVM) were utilized to evaluate the lipid content and inflammatory markers in mice and human liver tissue. NIR-HSI was also used to evaluate the degree of I/R injury and early allograft dysfunction (EAD) of liver transplant recipients enrolled in a retrospective cohort study.
RESULTS: NIR-HSI revealed significant differences in the severity of both fatty liver and I/R injury, specifically in the spectral range of 900-1700 nm. By integrating 1D-CNN into the spectral data analysis, we established a confusion matrix to assess the degrees of pathological changes in fatty liver and I/R injury, achieving high accuracy (all precision > 0.95) and F1 scores exceeding 0.90. Moreover, our model based on LSSVM demonstrated good normalized root mean square errors (NRMSE) for predicting total cholesterol, triglyceride, IL-1β, TNF-α, and HMGB1 levels (all NRMSE < 0.10), along with high coefficients of determination (all R2 > 0.9). Notably, NIR-HSI demonstrated a strong capacity to predict the presence of EAD following liver transplantation (LT). Targeted lipid metabolomics was used to explore the underlying mechanism of the association between spectral sensitive metabolites and metabolome.
CONCLUSIONS: Our findings highlight the potential of NIR-HSI enhanced by artificial intelligence in both basic and clinical research for assessing the severity of hepatic steatosis and I/R injury, predicting inflammatory marker levels, and identifying postoperative complication of EAD after LT.
MATERIAL AND METHODS: NIR-HSI data were used in conjunction with a one-dimensional convolutional neural network (1D-CNN) to classify the degree of hepatic steatosis and hepatic I/R injury. Additionally, least squares support vector machines (LSSVM) were utilized to evaluate the lipid content and inflammatory markers in mice and human liver tissue. NIR-HSI was also used to evaluate the degree of I/R injury and early allograft dysfunction (EAD) of liver transplant recipients enrolled in a retrospective cohort study.
RESULTS: NIR-HSI revealed significant differences in the severity of both fatty liver and I/R injury, specifically in the spectral range of 900-1700 nm. By integrating 1D-CNN into the spectral data analysis, we established a confusion matrix to assess the degrees of pathological changes in fatty liver and I/R injury, achieving high accuracy (all precision > 0.95) and F1 scores exceeding 0.90. Moreover, our model based on LSSVM demonstrated good normalized root mean square errors (NRMSE) for predicting total cholesterol, triglyceride, IL-1β, TNF-α, and HMGB1 levels (all NRMSE < 0.10), along with high coefficients of determination (all R2 > 0.9). Notably, NIR-HSI demonstrated a strong capacity to predict the presence of EAD following liver transplantation (LT). Targeted lipid metabolomics was used to explore the underlying mechanism of the association between spectral sensitive metabolites and metabolome.
CONCLUSIONS: Our findings highlight the potential of NIR-HSI enhanced by artificial intelligence in both basic and clinical research for assessing the severity of hepatic steatosis and I/R injury, predicting inflammatory marker levels, and identifying postoperative complication of EAD after LT.