Department of Radiation Oncology, Keck School of Medicine, University of Southern California, Los Angeles, CA, United States of America.
Department of Neurological Surgery, Keck School of Medicine, University of Southern California, Los Angeles, CA, United States of America.
Phys Med Biol. 2021 Jan 13;66(1):015003. doi: 10.1088/1361-6560/abca53.
Detection of brain metastases is a paramount task in cancer management due both to the number of high-risk patients and the difficulty of achieving consistent detection. In this study, we aim to improve the accuracy of automated brain metastasis (BM) detection methods using a novel asymmetric UNet (asym-UNet) architecture. An end-to-end asymmetric 3D-UNet architecture, with two down-sampling arms and one up-sampling arm, was constructed to capture the imaging features. The two down-sampling arms were trained using two different kernels (3 × 3 × 3 and 1 × 1 × 3, respectively) with the kernel (1 × 1 × 3) dominating the learning. As a comparison, vanilla single 3D UNets were trained with different kernels and evaluated using the same datasets. Voxel-based Dice similarity coefficient (DSC), sensitivity (S ), precision (P ), BM-based sensitivity (S ), and false detection rate (F ) were used to evaluate model performance. Contrast-enhanced T1 MR images from 195 patients with a total of 1034 BMs were solicited from our institutional stereotactic radiosurgery database. The patient cohort was split into training (160 patients, 809 lesions), validation (20 patients, 136 lesions), and testing (15 patients, 89 lesions) datasets. The lesions in the testing dataset were further divided into two subgroups based on the diameters (small S = 1-10 mm, large L = 11-26 mm). In the testing dataset, there were 72 and 17 BMs in the S and L sub-groups, respectively. Among all trained networks, asym-UNet achieved the highest DSC of 0.84 and lowest F of 0.24. Although vanilla 3D-UNet with a single 1 × 1 × 3 kernel achieved the highest sensitivities for the S group, it resulted in the lowest precision and highest false detection rate. Asym-UNet was shown to balance sensitivity and false detection rate as well as keep the segmentation accuracy high. The novel asym-UNet segmentation network showed overall competitive segmentation performance and more pronounced improvement in hard-to-detect small BMs comparing to the vanilla single 3D UNet.
脑转移瘤的检测在癌症管理中至关重要，这既是因为高危患者数量众多，也是因为难以实现一致的检测。在这项研究中，我们旨在通过一种新的非对称 U 型网络（asym-UNet）架构来提高自动脑转移瘤（BM）检测方法的准确性。构建了一个端到端的非对称 3D-U 型网络架构，具有两个下采样臂和一个上采样臂，用于捕获成像特征。两个下采样臂分别使用两个不同的核（3×3×3 和 1×1×3）进行训练，其中核（1×1×3）占据主导地位。作为对比，分别使用不同的核训练了普通的单 3D-U 型网络，并使用相同的数据集进行评估。基于体素的 Dice 相似系数（DSC）、灵敏度（S）、精度（P）、基于 BM 的灵敏度（S）和假阳性率（F）来评估模型性能。从我们的立体定向放射外科数据库中征集了 195 名患者共 1034 个 BM 的增强 T1MR 图像。将患者队列分为训练集（160 名患者，809 个病灶）、验证集（20 名患者，136 个病灶）和测试集（15 名患者，89 个病灶）。将测试集中的病灶进一步根据直径分为两个亚组（小 S = 1-10mm，大 L = 11-26mm）。在测试集中，小 S 组和大 L 组分别有 72 个和 17 个 BM。在所有训练好的网络中，asym-UNet 达到了 0.84 的最高 DSC 和 0.24 的最低 F。虽然具有单个 1×1×3 核的普通 3D-U 型网络对 S 组的灵敏度最高，但它导致了最低的精度和最高的假阳性率。asym-UNet 表现出平衡的灵敏度和假阳性率，并保持了较高的分割精度。与普通的单 3D-U 型网络相比，新的 asym-UNet 分割网络显示出整体具有竞争力的分割性能，并且在检测困难的小 BM 方面有更显著的改进。
