Grant: Alliance Grants

This research partnership has the potential to create new technologies that enhance the precision and reliability of localization and registration in total knee arthroplasty (TKA), which could improve surgical outcomes. In addition, it will train two highly qualified personnel (HQP) who will work closely with a Canadian company. To achieve these goals, this partnership has five main aims: 1) Estimate Points on Inaccessible Bone Regions with Deep Learning (DL): We will train a neural network to estimate points on parts of the bone that can't be directly accessed during surgery. These points can help reduce registration failures during surgery. 2) Upsample Points: We'll develop a method to add more points on the bone using pre-operative CT scans. This approach will help make the registration more accurate and efficient, saving the surgeon's time by requiring fewer points to be selected. 3) Estimate Points with Ultrasound: Instead of deep learning, we'll use ultrasound imaging to find points on bone regions that are hard to access. By segmenting ultrasound images, we can estimate point clouds from the bone surface, which will further help reduce registration failures. 4) Combine ultrasound and DL for Registration: We'll merge the points estimated from DL (Aims 1,2) and ultrasound (Aim 3) to provide a more reliable registration process. This combination will use the strengths of both methods to improve accuracy. 5) Imageless Registration Without CT: We'll explore if we can completely avoid the need for pre-operative CT scans. By using ultrasound to create 3D maps and fitting these maps to statistical bone models, we aim to reconstruct the bone accurately. These goals aim to make bone registration during surgery more accurate, reduce failures, and save time for surgeons by combining advanced DL and ultrasound technologies.