Institute of Biophotonics, National Yang Ming Chiao Tung University, Taipei City, Taiwan.
Department of Radiation Oncology, Taichung Veterans General Hospital, Taichung, Taiwan.
J Magn Reson Imaging. 2024 Jun;59(6):1967-1975. doi: 10.1002/jmri.28950. Epub 2023 Aug 12.
Deep learning-based segmentation algorithms usually required large or multi-institute data sets to improve the performance and ability of generalization. However, protecting patient privacy is a key concern in the multi-institutional studies when conventional centralized learning (CL) is used.
To explores the feasibility of a proposed lesion delineation for stereotactic radiosurgery (SRS) scheme for federated learning (FL), which can solve decentralization and privacy protection concerns.
Retrospective.
506 and 118 vestibular schwannoma patients aged 15-88 and 22-85 from two institutes, respectively; 1069 and 256 meningioma patients aged 12-91 and 23-85, respectively; 574 and 705 brain metastasis patients aged 26-92 and 28-89, respectively.
FIELD STRENGTH/SEQUENCE: 1.5T, spin-echo, and gradient-echo [Correction added after first online publication on 21 August 2023. Field Strength has been changed to "1.5T" from "5T" in this sentence.].
The proposed lesion delineation method was integrated into an FL framework, and CL models were established as the baseline. The effect of image standardization strategies was also explored. The dice coefficient was used to evaluate the segmentation between the predicted delineation and the ground truth, which was manual delineated by neurosurgeons and a neuroradiologist.
The paired t-test was applied to compare the mean for the evaluated dice scores (p < 0.05).
FL performed the comparable mean dice coefficient to CL for the testing set of Taipei Veterans General Hospital regardless of standardization and parameter; for the Taichung Veterans General Hospital data, CL significantly (p < 0.05) outperformed FL while using bi-parameter, but comparable results while using single-parameter. For the non-SRS data, FL achieved the comparable applicability to CL with mean dice 0.78 versus 0.78 (without standardization), and outperformed to the baseline models of two institutes.
The proposed lesion delineation successfully implemented into an FL framework. The FL models were applicable on SRS data of each participating institute, and the FL exhibited comparable mean dice coefficient to CL on non-SRS dataset. Standardization strategies would be recommended when FL is used.
4 TECHNICAL EFFICACY: Stage 1.
基于深度学习的分割算法通常需要大型或多机构数据集来提高性能和泛化能力。然而，当使用传统的集中式学习（CL）时，保护患者隐私是多机构研究中的一个关键问题。
探索一种用于联邦学习（FL）的立体定向放射外科（SRS）方案的提议病变勾画的可行性，该方案可以解决分散化和隐私保护问题。
回顾性。
分别来自两个机构的 506 名和 118 名年龄在 15-88 岁和 22-85 岁的前庭神经鞘瘤患者；分别来自两个机构的 1069 名和 256 名年龄在 12-91 岁和 23-85 岁的脑膜瘤患者；分别来自两个机构的 574 名和 705 名年龄在 26-92 岁和 28-89 岁的脑转移瘤患者。
磁场强度/序列：1.5T，自旋回波，梯度回波[在 2023 年 8 月 21 日首次在线发布后进行的更正。磁场强度已从“5T”更改为“1.5T”。]
所提出的病变勾画方法被集成到一个 FL 框架中，并建立了 CL 模型作为基线。还探索了图像标准化策略的效果。使用 Dice 系数评估预测勾画与由神经外科医生和神经放射科医生手动勾画的ground truth 之间的分割。
应用配对 t 检验比较评估 Dice 得分的平均值（p < 0.05）。
对于无论是否进行标准化和参数调整，FL 对台北荣民总医院的测试集的平均 Dice 系数都与 CL 相当；对于台中荣民总医院的数据，CL 显著（p < 0.05）优于使用双参数的 FL，但使用单参数时结果相当。对于非 SRS 数据，FL 的适用性与 CL 相当，平均 Dice 系数为 0.78 与 0.78（未经标准化），并且优于两个机构的基线模型。
所提出的病变勾画成功地实现到 FL 框架中。FL 模型适用于每个参与机构的 SRS 数据，并且在非 SRS 数据集上，FL 与 CL 的平均 Dice 系数相当。建议在使用 FL 时使用标准化策略。
4 技术功效：阶段 1。
