文献速递肿瘤 PET/CT

深度学习检测前列腺癌复发:18F-FACBC(氟环丁氨酸,Axumin®)正电子发射断层扫描

Deep learning detection of prostate cancer recurrence with 18F-FACBC (fluciclovine, Axumin®) positron emission tomography.
PubMed 文献 · Eur J Nucl Med Mol Imaging · 2020-06-17

该研究评估深度学习分类器基于肿瘤复发和/或转移的存在区分正常与异常18F-FACBC PET扫描的性能,纳入233例生化复发前列腺癌患者的251次连续扫描。基于切片的2D-CNN模型敏感度90.7%、特异度95.1%、AUC 0.971(p<0.001),优于基于病例的2D-CNN和3D-CNN模型。结论认为深度学习能准确分类盆腔异常18F-FACBC PET图像,单切片预测分类器性能优于基于病例的预测。

要点

  • 基于切片的2D-CNN模型敏感度90.7%、特异度95.1%、AUC 0.971(p<0.001)。
  • 基于病例的2D-CNN敏感度85.7%、特异度71.4%、AUC 0.750(p=0.013)。
  • 基于病例的3D-CNN敏感度71.4%、特异度71.4%、AUC 0.699(p=0.053)。

核心结论

  • 深度学习能准确分类生化复发前列腺癌患者的盆腔异常18F-FACBC PET图像。
  • 使用单切片预测的深度学习分类器性能优于基于病例的预测。

研究设计与样本

  • 研究纳入2017年9月至2019年6月间233例生化复发前列腺癌患者的251次连续18F-FACBC PET扫描,其中18例患者有2次扫描。
  • PET图像以临床报告为金标准标记为正常或异常。
  • 卷积神经网络模型采用两种架构训练:基于切片的2D-CNN(ResNet-50)使用单切片,基于病例的方法使用每幅PET图像一百个切片训练同一2D-CNN和3D-CNN(ResNet-14)。
  • 模型性能在独立测试数据集上评估。

主要发现

  • 基于切片的2D-CNN方法中,训练集使用6800个切片,测试集使用536个切片。
  • 基于切片的2D-CNN模型敏感度90.7%、特异度95.1%,受试者工作特征曲线下面积(AUC)为0.971(p<0.001)。
  • 基于病例的2D-CNN和3D-CNN架构方法中,随机分配训练集100幅图像、测试集28幅图像。
  • 基于病例的2D-CNN方法区分异常图像的敏感度85.7%、特异度71.4%、AUC 0.750(p=0.013)。
  • 基于病例的3D-CNN方法区分异常图像的敏感度71.4%、特异度71.4%、AUC 0.699(p=0.053)。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

PURPOSE: To evaluate the performance of deep learning (DL) classifiers in discriminating normal and abnormal 18F-FACBC (fluciclovine, Axumin®) PET scans based on the presence of tumor recurrence and/or metastases in patients with prostate cancer (PC) and biochemical recurrence (BCR). METHODS: A total of 251 consecutive 18F-fluciclovine PET scans were acquired between September 2017 and June 2019 in 233 PC patients with BCR (18 patients had 2 scans). PET images were labeled as normal or abnormal using clinical reports as the ground truth. Convolutional neural network (CNN) models were trained using two different architectures, a 2D-CNN (ResNet-50) using single slices (slice-based approach) and the same 2D-CNN and a 3D-CNN (ResNet-14) using a hundred slices per PET image (case-based approach). Models' performances were evaluated on independent test datasets. RESULTS: For the 2D-CNN slice-based approach, 6800 and 536 slices were used for training and test datasets, respectively. The sensitivity and specificity of this model were 90.7% and 95.1%, and the area under the curve (AUC) of receiver operating characteristic curve was 0.971 (p < 0.001). For the case-based approaches using both 2D-CNN and 3D-CNN architectures, a training dataset of 100 images and a test dataset of 28 images were randomly allocated. The sensitivity, specificity, and AUC to discriminate abnormal images by the 2D-CNN and 3D-CNN case-based approaches were 85.7%, 71.4%, and 0.750 (p = 0.013) and 71.4%, 71.4%, and 0.699 (p = 0.053), respectively. CONCLUSION: DL accurately classifies abnormal 18F-fluciclovine PET images of the pelvis in patients with BCR of PC. A DL classifier using single slice prediction had superior performance over case-based prediction.

来源:Eur J Nucl Med Mol Imaging,查看原文。版权归原作者所有。

引用

  1. Lee JJ, Yang H, Franc BL, et al. Deep learning detection of prostate cancer recurrence with 18F-FACBC (fluciclovine, Axumin®) positron emission tomography. Eur J Nucl Med Mol Imaging. 2020. PMID: 32556481. doi:10.1007/s00259-020-04912-w

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。