Swin UNETR segmentation with automated geometry filtering for biomechanical modeling of knee joint cartilage

Our study aimed to enhance subject-specific knee joint FE modeling by incorporating an automated knee cartilage segmentation algorithm. This segmentation was a 3D Swin UNETR for an initial segmentation of the femoral and tibial cartilages, followed by an automated filtering to improve surface roughness and continuity. Five hundred and seven magnetic resonance images (MRIs) from the Osteoarthritis Initiative (OAI) database were used to build and validate the segmentation model. The masks for cartilages were performed by skilled users from the Zuse Institute Berlin "F. Ambellan, A. Tack, M. Ehlke, and S. Zachow, “Automated segmentation of knee bone and cartilage combining statistical shape knowledge and convolutional neural networks: Data from the Osteoarthritis Initiative,” Med Image Anal, vol. 52, pp. 109–118, 2019." The Swin UNETR models and codes, and the filtering codes have been made publicly available, so researchers can use these models or customize the code for a different dataset to meet their needs. Please refer to our GitHub for any update https://github.com/Rezakaka/knee-segmentation.git本研究旨在通过引入自动化膝关节软骨分割算法，优化个性化膝关节有限元（Finite Element, FE）建模。该分割流程首先采用3D Swin UNETR对股骨及胫骨软骨执行初始分割，随后通过自动滤波操作优化表面粗糙度与连续性。本研究使用了来自骨关节炎倡议（Osteoarthritis Initiative, OAI）数据库的507例磁共振成像（Magnetic Resonance Imaging, MRI）数据，用于构建并验证该分割模型。膝关节软骨的掩码标注由柏林祖斯研究所的专业人员完成，相关标注方案参考了以下文献：F. Ambellan、A. Tack、M. Ehlke与S. Zachow发表于《医学图像分析（Medical Image Analysis）》2019年第52卷第109-118页的论文《结合统计形状知识与卷积神经网络的膝关节骨与软骨自动分割：来自骨关节炎倡议的数据集》。 本研究已公开Swin UNETR模型、配套代码及滤波代码，研究人员可直接使用这些模型，或针对不同数据集自定义代码以满足自身研究需求。如需获取最新更新，请访问本研究的GitHub仓库：https://github.com/Rezakaka/knee-segmentation.git