Department of Clinical Medicine and Surgery, University of Naples "Federico II", Naples, Italy.
Laboratory of Augmented Reality for Health Monitoring (ARHeMLab), Department of Electrical Engineering and Information Technology, University of Naples "Federico II", Naples, Italy.
J Magn Reson Imaging. 2021 Aug;54(2):452-459. doi: 10.1002/jmri.27585. Epub 2021 Feb 26.
Prostate volume, as determined by magnetic resonance imaging (MRI), is a useful biomarker both for distinguishing between benign and malignant pathology and can be used either alone or combined with other parameters such as prostate-specific antigen.
This study compared different deep learning methods for whole-gland and zonal prostate segmentation.
Retrospective.
A total of 204 patients (train/test = 99/105) from the PROSTATEx public dataset.
FIELD STRENGTH/SEQUENCE: A 3 T, TSE T -weighted.
Four operators performed manual segmentation of the whole-gland, central zone + anterior stroma + transition zone (TZ), and peripheral zone (PZ). U-net, efficient neural network (ENet), and efficient residual factorized ConvNet (ERFNet) were trained and tuned on the training data through 5-fold cross-validation to segment the whole gland and TZ separately, while PZ automated masks were obtained by the subtraction of the first two.
Networks were evaluated on the test set using various accuracy metrics, including the Dice similarity coefficient (DSC). Model DSC was compared in both the training and test sets using the analysis of variance test (ANOVA) and post hoc tests. Parameter number, disk size, training, and inference times determined network computational complexity and were also used to assess the model performance differences. A P < 0.05 was selected to indicate the statistical significance.
The best DSC (P < 0.05) in the test set was achieved by ENet: 91% ± 4% for the whole gland, 87% ± 5% for the TZ, and 71% ± 8% for the PZ. U-net and ERFNet obtained, respectively, 88% ± 6% and 87% ± 6% for the whole gland, 86% ± 7% and 84% ± 7% for the TZ, and 70% ± 8% and 65 ± 8% for the PZ. Training and inference time were lowest for ENet.
Deep learning networks can accurately segment the prostate using T -weighted images.
4 TECHNICAL EFFICACY: Stage 2.
通过磁共振成像（MRI）确定的前列腺体积是区分良性和恶性病理的有用生物标志物，可单独使用或与前列腺特异性抗原等其他参数结合使用。
本研究比较了用于全腺和分区前列腺分割的不同深度学习方法。
回顾性。
来自 PROSTATEx 公共数据集的 204 名患者（训练/测试=99/105）。
磁场强度/序列：3T，TSE T1 加权。
四名操作员对全腺、中央区+前基质+移行区（TZ）和周围区（PZ）进行了手动分割。U-net、高效神经网络（ENet）和高效残差因子化 ConvNet（ERFNet）通过 5 折交叉验证在训练数据上进行训练和调整，以分别分割全腺和 TZ，而通过减去前两个区域获得 PZ 的自动掩模。
使用各种准确性指标（包括 Dice 相似系数（DSC））在测试集上评估网络。使用方差分析（ANOVA）和事后检验比较了模型在训练集和测试集上的 DSC。网络的参数数量、磁盘大小、训练和推理时间决定了网络的计算复杂度，也用于评估模型性能差异。选择 P<0.05 表示具有统计学意义。
在测试集中，ENet 实现了最佳的 DSC（P<0.05）：全腺为 91%±4%，TZ 为 87%±5%，PZ 为 71%±8%。U-net 和 ERFNet 分别获得全腺 88%±6%和 87%±6%，TZ 为 86%±7%和 84%±7%，PZ 为 70%±8%和 65%±8%。ENet 的训练和推理时间最短。
深度学习网络可以使用 T1 加权图像准确分割前列腺。
4 级技术功效：2 级。
