Fukuda Norio, Otake Yoshito, Takao Masaki, Yokota Futoshi, Ogawa Takeshi, Uemura Keisuke, Nakaya Ryota, Tamura Kazunori, Grupp Robert B, Farvardin Amirhossein, Armand Mehran, Sugano Nobuhiko, Sato Yoshinobu
Graduate School of Information Science, Nara Institute of Science and Technology, Ikoma, Nara, Japan.
Graduate School of Medicine, Osaka University, Suita, Osaka, Japan.
Int J Comput Assist Radiol Surg. 2017 May;12(5):733-742. doi: 10.1007/s11548-016-1519-8. Epub 2017 Feb 10.
Patient-specific musculoskeletal biomechanical simulation is useful in preoperative surgical planning and postoperative assessment in orthopedic surgery and rehabilitation medicine. A difficulty in application of the patient-specific musculoskeletal modeling comes from the fact that the muscle attachment regions are typically invisible in CT and MRI. Our purpose is to develop a method for estimating patient-specific muscle attachment regions from 3D medical images and to validate with cadaver experiments.
Eight fresh cadaver specimens of the lower extremity were used in the experiments. Before dissection, CT images of all the specimens were acquired and the bone regions in CT images were extracted using an automated segmentation method to reconstruct the bone shape models. During dissection, ten different muscle attachment regions were recorded with an optical motion tracker. Then, these regions obtained from eight cadavers were integrated on an average bone surface via non-rigid registration, and muscle attachment probabilistic atlases (PAs) were constructed. An average muscle attachment region derived from the PA was non-rigidly mapped to the patients bone surface to estimate the patient-specific muscle attachment region.
Average Dice similarity coefficient between the true and estimated attachment areas computed by the proposed method was more than 10% higher than the one computed by a previous method in most cases and the average boundary distance error of the proposed method was 1.1 mm smaller than the previous method on average.
We conducted cadaver experiments to measure the attachment regions of the hip muscles and constructed PAs of the muscle attachment regions. The muscle attachment PA clarified the variations of the location of the muscle attachments and allowed us to estimate the patient-specific attachment area more accurately based on the patient bone shape derived from CT.
针对患者的肌肉骨骼生物力学模拟在骨科手术和康复医学的术前手术规划及术后评估中很有用。患者特异性肌肉骨骼建模应用中的一个难点在于，肌肉附着区域在CT和MRI中通常不可见。我们的目的是开发一种从3D医学图像估计患者特异性肌肉附着区域的方法，并通过尸体实验进行验证。
实验使用了8个新鲜的下肢尸体标本。在解剖前，获取所有标本的CT图像，并使用自动分割方法提取CT图像中的骨骼区域以重建骨骼形状模型。在解剖过程中，用光学运动跟踪器记录10个不同的肌肉附着区域。然后，通过非刚性配准将从8具尸体获得的这些区域整合到平均骨骼表面上，并构建肌肉附着概率图谱（PAs）。将从PA得出的平均肌肉附着区域非刚性映射到患者的骨骼表面，以估计患者特异性肌肉附着区域。
在大多数情况下，通过本方法计算得到的真实附着区域与估计附着区域之间的平均骰子相似系数比先前方法计算得到的高出10%以上，并且本方法的平均边界距离误差平均比先前方法小1.1毫米。
我们进行了尸体实验以测量髋部肌肉的附着区域，并构建了肌肉附着区域的PAs。肌肉附着PA阐明了肌肉附着位置的变化，并使我们能够基于从CT得出的患者骨骼形状更准确地估计患者特异性附着区域。
