oa Feasibility study on MRI segmentation of knee structures for computer-assisted surgery

BACKGROUND AND OBJECTIVE: Knee surgeries for total knee replacement or Anterior Cruciate Ligament (ACL) repair involve the use of intraoperative computer-assisted navigation techniques for guiding surgical tools during the procedure. In clinical practice, a digital map of the knee is created for navigation from images acquired preoperatively. High hard-tissue contrast makes CT the preferred imaging modality. However certain knee surgeries, such as ACL repair, require segmentation of soft-tissue structures for navigation. As opposed to CT, MRI has soft-tissue contrast. Considering this advantage, we investigate the applicability of MRI for segmenting both hard-tissue as well as soft-tissue structures in ACL repair surgery. METHODS: MR images (3D-DESS protocol; Pixel Size = 0.46x0.46mm²; FoV = 150x150mm²; slice thickness = 3mm; inter-slice distance = 3mm) of the knee at four flexion positions (30°, 45°, 90°, Full-Extension) were collected using Siemens Espree scanner on eight healthy volunteers. Two experts manually delineated the MR images using OsiriX software. The delineated boundaries of the hard-tissues (Tibia, Femur, and Patella) and soft-tissue (ACL) were used to create respective binary masks which were then combined into a single mask for each tissue type using STAPLE algorithm. The output was fed to Marching-Cube algorithm followed by Laplacian smoothing filter to generate triangular meshes of the knee structures. These 3D meshes can be used as a digital map to ease the navigation (Figure 1). RESULTS: The binary mask overlap between experts (Table 1) is used to measure the reproducibility of the segmentation and hence the confidence in generating 3D models for navigation. The high overlap for hard-tissues shows that MRI is relevant to segment them. The lower overlap for soft-tissues shows a perfectible segmentation, but still sufficient to show that their segmentation is achievable using MRI. CONCLUSIONS: This work is a first step towards using preoperative MRI segmentations for surgical navigation in knee surgeries showing it can provide not only hard- but also soft-tissue information as compared to CT.