Academic radiology · 2026
An Interpretable CT-based Prediction Model Assists in Predicting Length of Stay in Ulcerative Colitis Patients: A Multicenter Study.
СтатьяМногоцентровое исследованиеНазвание 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.
Lu Jun, Zhao Hongzhe, Li Junlin, Xu Hui, Zhang Jingxuan, Cheng Tianxin, Guo Jia, Han Xinjun, Wang Yuxin, Meng Xuxu, Jiang Jiahui, Sun Haoran, Wang Zhenchang, Xu Lixue, Cui Jianmin, Yang Zhenghan
Аннотация
RATIONALE AND OBJECTIVES: This study aimed to develop and validate an interpretable prediction model integrating CT-derived imaging biomarkers and clinical laboratory parameters to predict prolonged length of stay (LOS) in ulcerative colitis (UC).
MATERIALS AND METHODS: In this multi-center retrospective study, 273 UC patients were enrolled from three hospitals. Patients were divided into training (n = 184) and external validation (n = 89) cohorts. CT visual features and body composition parameters were evaluated. Univariate and multivariable logistic regression were used to select predictive features. The imaging, clinical, and combined model were developed and evaluated. The feature contribution was qualified using the Shapley additive explanation (SHAP) algorithm.
RESULTS: The combined model with six key predictors including visceral adipose tissue (VAT) density, visceral obesity, bowel wall thickening, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR) and albumin demonstrated superior performance with area under the curve (AUCs) of 0.862 and 0.867 in the training and validation cohorts for predicting prolonged LOS(LOS≥14 days), respectively. It significantly outperformed both the imaging-only and clinical-only models (all p < 0.05). SHAP analysis confirmed VAT density as the most influential feature. With the addition of the clinical biomarkers to the CT-derived prediction model, the prediction performance significantly improved with integrated discrimination improvement (IDI=0.187, 0.189) and net reclassification improvement (NRI=0.193, 0.268) in the training and validation cohort, respectively (all p < 0.05).
CONCLUSION: The combined prediction model is promising for predicting prolonged LOS in UC. Moreover, VAT density was established as an independent predictor of LOS and was identified by SHAP analysis as the most influential feature in the predictive model.
MATERIALS AND METHODS: In this multi-center retrospective study, 273 UC patients were enrolled from three hospitals. Patients were divided into training (n = 184) and external validation (n = 89) cohorts. CT visual features and body composition parameters were evaluated. Univariate and multivariable logistic regression were used to select predictive features. The imaging, clinical, and combined model were developed and evaluated. The feature contribution was qualified using the Shapley additive explanation (SHAP) algorithm.
RESULTS: The combined model with six key predictors including visceral adipose tissue (VAT) density, visceral obesity, bowel wall thickening, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR) and albumin demonstrated superior performance with area under the curve (AUCs) of 0.862 and 0.867 in the training and validation cohorts for predicting prolonged LOS(LOS≥14 days), respectively. It significantly outperformed both the imaging-only and clinical-only models (all p < 0.05). SHAP analysis confirmed VAT density as the most influential feature. With the addition of the clinical biomarkers to the CT-derived prediction model, the prediction performance significantly improved with integrated discrimination improvement (IDI=0.187, 0.189) and net reclassification improvement (NRI=0.193, 0.268) in the training and validation cohort, respectively (all p < 0.05).
CONCLUSION: The combined prediction model is promising for predicting prolonged LOS in UC. Moreover, VAT density was established as an independent predictor of LOS and was identified by SHAP analysis as the most influential feature in the predictive model.