基于深度学习单发探测器的 T1 加权 MRI 立体定向放射手术中脑转移瘤的计算机辅助检测。
Computer-aided Detection of Brain Metastases in T1-weighted MRI for Stereotactic Radiosurgery Using Deep Learning Single-Shot Detectors.
机构信息
From the Department of Imaging Physics (Z.Z., J.W.S., J.B.S., J.M.), Medical Physics Graduate Program, UTHealth Graduate School of Biomedical Sciences (J.W.S.), Department of Diagnostic Radiology (J.M.J., M.K.G., M.M.C.), Department of Radiation Physics (T.M.B.), Department of Radiation Oncology (Y.W., J.L.), and Department of Cancer Systems Imaging (M.D.P.), The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Houston, TX 77030.
出版信息
Radiology. 2020 May;295(2):407-415. doi: 10.1148/radiol.2020191479. Epub 2020 Mar 17.
Background Brain metastases are manually identified during stereotactic radiosurgery (SRS) treatment planning, which is time consuming and potentially challenging. Purpose To develop and investigate deep learning (DL) methods for detecting brain metastasis with MRI to aid in treatment planning for SRS. Materials and Methods In this retrospective study, contrast material-enhanced three-dimensional T1-weighted gradient-echo MRI scans from patients who underwent gamma knife SRS from January 2011 to August 2018 were analyzed. Brain metastases were manually identified and contoured by neuroradiologists and treating radiation oncologists. DL single-shot detector (SSD) algorithms were constructed and trained to map axial MRI slices to a set of bounding box predictions encompassing metastases and associated detection confidences. Performances of different DL SSDs were compared for per-lesion metastasis-based detection sensitivity and positive predictive value (PPV) at a 50% confidence threshold. For the highest-performing model, detection performance was analyzed by using free-response receiver operating characteristic analysis. Results Two hundred sixty-six patients (mean age, 60 years ± 14 [standard deviation]; 148 women) were randomly split into 80% training and 20% testing groups (212 and 54 patients, respectively). For the testing group, sensitivity of the highest-performing (baseline) SSD was 81% (95% confidence interval [CI]: 80%, 82%; 190 of 234) and PPV was 36% (95% CI: 35%, 37%; 190 of 530). For metastases measuring at least 6 mm, sensitivity was 98% (95% CI: 97%, 99%; 130 of 132) and PPV was 36% (95% CI: 35%, 37%; 130 of 366). Other models (SSD with a ResNet50 backbone, SSD with focal loss, and RetinaNet) yielded lower sensitivities of 73% (95% CI: 72%, 74%; 171 of 234), 77% (95% CI: 76%, 78%; 180 of 234), and 79% (95% CI: 77%, 81%; 184 of 234), respectively, and lower PPVs of 29% (95% CI: 28%, 30%; 171 of 581), 26% (95% CI: 26%, 26%; 180 of 681), and 13% (95% CI: 12%, 14%; 184 of 1412). Conclusion Deep-learning single-shot detector models detected nearly all brain metastases that were 6 mm or larger with limited false-positive findings using postcontrast T1-weighted MRI. © RSNA, 2020 See also the editorial by Kikinis and Wells in this issue.
背景 脑转移瘤在立体定向放射外科（SRS）治疗计划中需要手动识别，这既耗时又具有挑战性。目的 开发并研究用于 MRI 检测脑转移瘤的深度学习（DL）方法，以辅助 SRS 治疗计划。材料与方法 本回顾性研究分析了 2011 年 1 月至 2018 年 8 月期间接受伽玛刀 SRS 的患者的对比增强三维 T1 加权梯度回波 MRI 扫描。由神经放射科医生和治疗放射肿瘤学家手动识别和勾画脑转移瘤。构建并训练了深度学习单发探测器（SSD）算法，以将轴向 MRI 切片映射到一组包含转移瘤及其相关检测置信度的边界框预测。比较了不同深度学习 SSD 的性能，以基于病变的转移瘤检测灵敏度和 50%置信度阈值的阳性预测值（PPV）。对于性能最高的模型，使用自由响应接收器操作特征分析来分析检测性能。结果 在 266 名患者（平均年龄，60 岁±14[标准差]；148 名女性）中，随机分为 80%的训练组和 20%的测试组（分别为 212 名和 54 名患者）。对于测试组，性能最高的（基线）SSD 的灵敏度为 81%（95%置信区间：80%，82%；234 例中有 190 例），PPV 为 36%（95%置信区间：35%，37%；530 例中有 190 例）。对于至少 6mm 的转移瘤，灵敏度为 98%（95%置信区间：97%，99%；132 例中有 130 例），PPV 为 36%（95%置信区间：35%，37%；366 例中有 130 例）。其他模型（基于 ResNet50 骨干的 SSD、基于焦点损失的 SSD 和 RetinaNet）的灵敏度分别为 73%（95%置信区间：72%，74%；234 例中有 171 例）、77%（95%置信区间：76%，78%；234 例中有 180 例）和 79%（95%置信区间：77%，81%；234 例中有 184 例），PPV 分别为 29%（95%置信区间：28%，30%；581 例中有 171 例）、26%（95%置信区间：26%，26%；681 例中有 180 例）和 13%（95%置信区间：12%，14%；1412 例中有 184 例）。结论 使用增强 T1 加权 MRI，深度学习单发探测器模型可以检测出几乎所有大小为 6mm 或以上的脑转移瘤，假阳性发现有限。©RSNA，2020 另请参阅本期 Kikinis 和 Wells 的社论。
Zotero 插件