Computer Vision News - August 2022

41 AI to support THR registration or 3D tool tracking using the images from the operating room (OR), or other tracking equipment. Dedicated AI algorithms can register the live image with the pre-op scan and using augmented reality the tool position can be overlayed on the surgical plan. Preparing the bone for implant insertion requires removal of the diseased bone and cartilage. Removing too much or too little can reduce procedural success rates. Roboticcontrolof thetools incombination with accurate registration of the surgical scene with the pre-op plan, prevents any alterations from the original plan . One of the most difficult steps in THR is implanting the stem in the correct angle in and assesses the proper approach for the surgery.Recentadvancementsincomputer- vision (CV) and deep-learning (DL) allow accurate, automatic segmentation of the hip and femur bones from either/all imaging modality, and reconstructing a 3D model from CT, MRI, and even from 2D X-ray images . This 3D model can be used to test the fitting of implants (i.e., need for high offset), calculate bone- removal quantity needed for proper fitting, and recover any other patient-specific information (i.e. need for specific implant design). This model can be also used for 3D printing for hands-on practice. SpecificMRI protocols allow soft-tissue segmentation for cases which require more attention and unique planning (obese patients, soft- tissue malformations, etc.). If the procedure is conducted with the assistance of a robot, a specific surgical plan can be prepared and uploaded to the robot . All procedural steps will be guided, and any deviation from the plan will be detected by the robot and the surgeon will be alerted. Real-time guidance To verify the plan proceeds accordingly, real-time tracking and registration of the surgical tools is required. This can be achieved using manual