Department of Medical Biophysics, University of Toronto, Toronto, Canada.
Princess Margaret Cancer Center, University Health Network, Toronto, CA, Canada; Department of Radiation Oncology, University of Toronto, Toronto, CA, Canada.
Radiother Oncol. 2024 Aug;197:110332. doi: 10.1016/j.radonc.2024.110332. Epub 2024 May 18.
Deep learning can automate delineation in radiation therapy, reducing time and variability. Yet, its efficacy varies across different institutions, scanners, or settings, emphasizing the need for adaptable and robust models in clinical environments. Our study demonstrates the effectiveness of the transfer learning (TL) approach in enhancing the generalizability of deep learning models for auto-segmentation of organs-at-risk (OARs) in cervical brachytherapy.
A pre-trained model was developed using 120 scans with ring and tandem applicator on a 3T magnetic resonance (MR) scanner (RT3). Four OARs were segmented and evaluated. Segmentation performance was evaluated by Volumetric Dice Similarity Coefficient (vDSC), 95 % Hausdorff Distance (HD95), surface DSC, and Added Path Length (APL). The model was fine-tuned on three out-of-distribution target groups. Pre- and post-TL outcomes, and influence of number of fine-tuning scans, were compared. A model trained with one group (Single) and a model trained with all four groups (Mixed) were evaluated on both seen and unseen data distributions.
TL enhanced segmentation accuracy across target groups, matching the pre-trained model's performance. The first five fine-tuning scans led to the most noticeable improvements, with performance plateauing with more data. TL outperformed training-from-scratch given the same training data. The Mixed model performed similarly to the Single model on RT3 scans but demonstrated superior performance on unseen data.
TL can improve a model's generalizability for OAR segmentation in MR-guided cervical brachytherapy, requiring less fine-tuning data and reduced training time. These results provide a foundation for developing adaptable models to accommodate clinical settings.
深度学习可实现放射治疗中的自动勾画，从而减少时间和变异性。然而，其在不同机构、扫描仪或环境中的效果存在差异，这强调了在临床环境中需要适应性强且稳健的模型。我们的研究表明，迁移学习（TL）方法可有效提高深度学习模型对磁共振引导下宫颈癌近距离放疗中危及器官（OAR）自动勾画的泛化能力。
使用带有环和 tandem 施源器的 120 个 3T 磁共振（MR）扫描仪（RT3）上的扫描数据来开发预训练模型。对四个 OAR 进行分割和评估。通过体积 Dice 相似系数（vDSC）、95%Hausdorff 距离（HD95）、表面 DSC 和附加路径长度（APL）评估分割性能。在三个非分布目标组上对模型进行微调。比较了微调前后的结果以及微调扫描数量的影响。在已有和新的数据集上评估了仅使用一个组（Single）和使用所有四个组（Mixed）进行训练的模型。
TL 提高了跨目标组的分割准确性，与预训练模型的性能相匹配。前五个微调扫描带来了最显著的改进，随着数据量的增加，性能趋于平稳。与使用相同训练数据的从头开始训练相比，TL 表现更好。在 RT3 扫描上，Mixed 模型的性能与 Single 模型相似，但在未见数据上表现更优。
TL 可提高 OAR 分割在 MR 引导下宫颈癌近距离放疗中的模型泛化能力，所需的微调数据更少，训练时间更短。这些结果为开发适应临床环境的适应性模型提供了基础。
