Gastro hep advances · 2025
Development of a Neural Network to Detect Hepatic Steatosis in Metabolic Dysfunction-Associated Steatotic Liver Disease.
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.
Hirooka Masashi, Miyake Teruki, Yano Ryo, Nakamura Yoshiko, Okazaki Yuki, Shimamoto Toyoki, Yamamoto Yasunori, Watanabe Takao, Yoshida Osamu, Hirooka Kana, Tokumoto Yoshio, Abe Masanori, Iwata Takeru, Hiasa Yoichi
Аннотация
BACKGROUND AND AIMS: Early identification of metabolic dysfunction-associated steatotic liver disease (MASLD) is critical for risk stratification and timely intervention. Conventional noninvasive indices (fatty liver index, hepatic steatosis index, Zhejiang University index, and MASLD index) are limited by linear assumptions and moderate predictive accuracy. We aimed to develop and externally validate a neural network model for noninvasive detection of hepatic steatosis.
METHODS: We used a retrospective cohort of 17,465 Japanese health checkup participants (2010-2020). All clinical data were obtained at the Ehime General Health Care Association. The cohort was split into a development cohort (n = 8426) and an internal validation cohort (n = 9039). A feedforward neural network was trained using clinical and biochemical variables, including body mass index, abdominal circumference, and existing indices. External validation used the Third National Health and Nutrition Examination Survey cohort (n = 9759), with hepatic steatosis defined by ultrasonography (Gallbladder and Upper Abdominal Ultrasound Hepatic Steatosis Profile Rating≥2). Model performance was assessed via area under the receiver operating characteristic curve, calibration, decision curve analysis, and subgroup analyses.
RESULTS: The neural network achieved an area under the receiver operating characteristic curve of 0.922 (95% confidence interval: 0.913-0.931) in internal validation and 0.924 (95% confidence interval: 0.917-0.931) in external validation, outperforming fatty liver index, hepatic steatosis index, Zhejiang University index, and MASLD indices (all P < .001). At the optimal Youden cutoff (0.357), sensitivity and specificity were 89% and 86%, respectively. Calibration analysis and decision curve analysis confirmed strong agreement between predicted and observed risk. Abdominal circumference and body mass index were the most influential predictors.
CONCLUSION: Our neural network model outperforms conventional indices in detecting moderate-to-severe hepatic steatosis and may facilitate early, scalable MASLD screening in primary care and low-resource settings.UMIN No. 11953.
METHODS: We used a retrospective cohort of 17,465 Japanese health checkup participants (2010-2020). All clinical data were obtained at the Ehime General Health Care Association. The cohort was split into a development cohort (n = 8426) and an internal validation cohort (n = 9039). A feedforward neural network was trained using clinical and biochemical variables, including body mass index, abdominal circumference, and existing indices. External validation used the Third National Health and Nutrition Examination Survey cohort (n = 9759), with hepatic steatosis defined by ultrasonography (Gallbladder and Upper Abdominal Ultrasound Hepatic Steatosis Profile Rating≥2). Model performance was assessed via area under the receiver operating characteristic curve, calibration, decision curve analysis, and subgroup analyses.
RESULTS: The neural network achieved an area under the receiver operating characteristic curve of 0.922 (95% confidence interval: 0.913-0.931) in internal validation and 0.924 (95% confidence interval: 0.917-0.931) in external validation, outperforming fatty liver index, hepatic steatosis index, Zhejiang University index, and MASLD indices (all P < .001). At the optimal Youden cutoff (0.357), sensitivity and specificity were 89% and 86%, respectively. Calibration analysis and decision curve analysis confirmed strong agreement between predicted and observed risk. Abdominal circumference and body mass index were the most influential predictors.
CONCLUSION: Our neural network model outperforms conventional indices in detecting moderate-to-severe hepatic steatosis and may facilitate early, scalable MASLD screening in primary care and low-resource settings.UMIN No. 11953.