Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, Zurich, Switzerland; Department of Nuclear Medicine, University Hospital Zurich, Zurich, Switzerland; University of Zurich, Zurich, Switzerland.
University of Zurich, Zurich, Switzerland; Department of Surgery and Transplantation, University Hospital Zurich, Zurich, Switzerland; Department of Hepatobiliary Surgery and Liver Transplantation, St. Vincent's University Hospital, Dublin, Ireland.
Eur J Radiol. 2021 Jul;140:109733. doi: 10.1016/j.ejrad.2021.109733. Epub 2021 Apr 24.
To evaluate whether a magnetic resonance imaging (MRI) radiomics-based machine learning classifier can predict postoperative pancreatic fistula (POPF) after pancreaticoduodenectomy (PD) and to compare its performance to T1 signal intensity ratio (T1 SIratio).
Sixty-two patients who underwent 3 T MRI before PD between 2008 and 2018 were retrospectively analyzed. POPF was graded and split into clinically relevant POPF (CR-POPF) vs. biochemical leak or no POPF. On T1- and T2-weighted images, 2 regions of interest were placed in the pancreatic corpus and cauda. 173 radiomics features were extracted using pyRadiomics. Additionally, the pancreas-to-muscle T1 SIratio was measured. The dataset was augmented and split into training (70 %) and test sets (30 %). A Boruta algorithm was used for feature reduction. For prediction of CR-POPF models were built using a gradient-boosted tree (GBT) and logistic regression from the radiomics features, T1 SIratio and a combination of the two. Diagnostic accuracy of the models was compared using areas under the receiver operating characteristics curve (AUCs).
Five most important radiomics features were identified for prediction of CR-POPF. A GBT using these features achieved an AUC of 0.82 (95 % Confidence Interval [CI]: 0.74 - 0.89) when applied on the original (non-augmented) dataset. Using T1 SIratio, a GBT model resulted in an AUC of 0.75 (CI: 0.63 - 0.84) and a logistic regression model delivered an AUC of 0.75 (CI: 0.63 - 0.84). A GBT model combining radiomics features and T1 SIratio resulted in an AUC of 0.90 (CI 0.84 - 0.95).
MRI-radiomics with routine sequences provides promising prediction of CR-POPF.
评估基于磁共振成像（MRI）放射组学的机器学习分类器是否可用于预测胰十二指肠切除术（PD）后胰瘘（POPF），并比较其与 T1 信号强度比（T1 SIratio）的性能。
回顾性分析了 2008 年至 2018 年间在 3T MRI 前接受 PD 的 62 例患者。将 POPF 分级，并分为临床相关 POPF（CR-POPF）与生化漏或无 POPF。在 T1 加权和 T2 加权图像上，在胰腺体部和尾部放置 2 个感兴趣区。使用 pyRadiomics 提取 173 个放射组学特征。此外，还测量了胰腺与肌肉的 T1 SIratio。对数据集进行扩充并分为训练集（70％）和测试集（30％）。使用 Boruta 算法进行特征降维。使用梯度提升树（GBT）和逻辑回归，基于放射组学特征、T1 SIratio 以及两者的组合，构建预测 CR-POPF 的模型。使用受试者工作特征曲线下的面积（AUCs）比较模型的诊断准确性。
确定了用于预测 CR-POPF 的 5 个最重要的放射组学特征。在原始（未扩充）数据集上应用包含这些特征的 GBT，AUC 为 0.82（95％置信区间 [CI]：0.74 - 0.89）。使用 T1 SIratio，GBT 模型的 AUC 为 0.75（CI：0.63 - 0.84），逻辑回归模型的 AUC 为 0.75（CI：0.63 - 0.84）。结合放射组学特征和 T1 SIratio 的 GBT 模型的 AUC 为 0.90（CI 0.84 - 0.95）。
基于常规序列的 MRI 放射组学对 CR-POPF 的预测具有很大的潜力。
