Division of Hepatobiliary and Pancreatic Surgery, Department of Surgery, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea.
Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea.
Int J Surg. 2022 Sep;105:106851. doi: 10.1016/j.ijsu.2022.106851. Epub 2022 Aug 29.
Pancreatic ductal adenocarcinoma (PDAC) has a poor prognosis even after curative resection. A deep learning-based stratification of postoperative survival in the preoperative setting may aid the treatment decisions for improving prognosis. This study was aimed to develop a deep learning model based on preoperative data for predicting postoperative survival.
The patients who underwent surgery for PDAC between January 2014 and May 2015. Clinical data-based machine learning models and computed tomography (CT) data-based deep learning models were developed separately, and ensemble learning was utilized to combine two models. The primary outcomes were the prediction of 2-year overall survival (OS) and 1-year recurrence-free survival (RFS). The model's performance was measured by area under the receiver operating curve (AUC) and was compared with that of American Joint Committee on Cancer (AJCC) 8th stage.
The median OS and RFS were 23 and 10 months in training dataset (n = 229), and 22 and 11 months in test dataset (n = 53), respectively. The AUC of the ensemble model for predicting 2-year OS and 1-year RFS in the test dataset was 0.76 and 0.74, respectively. The performance of the ensemble model was comparable to that of the AJCC in predicting 2-year OS (AUC, 0.67; P = 0.35) and superior to the AJCC in predicting 1-year RFS (AUC, 0.54; P = 0.049).
Our ensemble model based on routine preoperative variables showed good performance for predicting prognosis for PDAC patients after surgery.
即使在根治性切除术后，胰腺导管腺癌（PDAC）的预后仍然很差。术前基于深度学习的术后生存分层可能有助于改善预后的治疗决策。本研究旨在开发一种基于术前数据的预测术后生存的深度学习模型。
该研究纳入了 2014 年 1 月至 2015 年 5 月期间接受 PDAC 手术的患者。分别建立了基于临床数据的机器学习模型和基于计算机断层扫描（CT）数据的深度学习模型，并利用集成学习将两种模型相结合。主要结局为预测 2 年总生存（OS）和 1 年无复发生存（RFS）。通过接受者操作特征曲线下面积（AUC）来衡量模型的性能，并与美国癌症联合委员会（AJCC）第 8 版进行比较。
在训练数据集（n=229）中，中位 OS 和 RFS 分别为 23 个月和 10 个月，在测试数据集（n=53）中，中位 OS 和 RFS 分别为 22 个月和 11 个月。在测试数据集，该集成模型预测 2 年 OS 和 1 年 RFS 的 AUC 分别为 0.76 和 0.74。该集成模型在预测 2 年 OS 方面的性能与 AJCC 相当（AUC，0.67；P=0.35），在预测 1 年 RFS 方面优于 AJCC（AUC，0.54；P=0.049）。
基于常规术前变量的我们的集成模型在预测 PDAC 患者手术后的预后方面表现良好。
