HPB : the official journal of the International Hepato Pancreato Biliary Association · 2026
Deep learning via pathology-image integration for margin prediction after pancreatoduodenectomy.
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.
Ricker Ansley B, Satyanarayana Sagar, Pritt Trenton, Jensen Stephanie, Peterson Katheryn, Mantha Rohit, Cochran Allyson, Martinie John B, Baker Erin H
Аннотация
INTRODUCTION: R1 margin status after pancreatoduodenectomy (PD) for pancreatic ductal adenocarcinoma (PDAC) remains common and is associated with inferior survival. Preoperative imaging inadequately predicts microscopic margin involvement, particularly at the superior mesenteric artery interface. This study evaluated whether deep learning applied to preoperative CT imaging could predict pathological margin status.
METHODS: Patients who underwent PD for PDAC between 2007-2021 were identified from a prospectively maintained database. Preoperative triphasic CT scans were processed into standardized axial slices and labeled according to pathological margin status. Two convolutional neural network models were developed using an attention-based multiple instance learning approach. Model 1 used all slices from each CT scan. Model 2 incorporated noise-reduction techniques, including selective slice extraction, grayscale normalization, artifact removal, and cropping to the pancreas and surrounding tissue.
RESULTS: One hundred patients met inclusion criteria (50 R0, 50 R1/R2), yielding 14,445 CT slices. Model 1 achieved an AUC of 0.69. After noise reduction, Model 2 demonstrated improved performance with an AUC of 0.78 and 88.2% sensitivity for identifying margin-positive cases.
CONCLUSION: Deep learning applied to preoperative CT imaging shows feasibility for predicting pathological margin status after PD. Noise-reduction preprocessing improved performance, supporting imaging in preoperative risk stratification and surgical planning.
METHODS: Patients who underwent PD for PDAC between 2007-2021 were identified from a prospectively maintained database. Preoperative triphasic CT scans were processed into standardized axial slices and labeled according to pathological margin status. Two convolutional neural network models were developed using an attention-based multiple instance learning approach. Model 1 used all slices from each CT scan. Model 2 incorporated noise-reduction techniques, including selective slice extraction, grayscale normalization, artifact removal, and cropping to the pancreas and surrounding tissue.
RESULTS: One hundred patients met inclusion criteria (50 R0, 50 R1/R2), yielding 14,445 CT slices. Model 1 achieved an AUC of 0.69. After noise reduction, Model 2 demonstrated improved performance with an AUC of 0.78 and 88.2% sensitivity for identifying margin-positive cases.
CONCLUSION: Deep learning applied to preoperative CT imaging shows feasibility for predicting pathological margin status after PD. Noise-reduction preprocessing improved performance, supporting imaging in preoperative risk stratification and surgical planning.