Bottlenecks in orthopedic procedures span across each of its primary steps, namely examination, planning and operation. Incorporation of new visualization technologies and planning methodologies have been shown to shorten the aforementioned steps, while retaining the high standard of accuracy that are required in these common practices. Innovative algorithmic techniques, relying on image processing, computer vision and machine learning are increasingly utilized in practice and have been gaining approval by regulatory bodies such as the FDA, into what is better known as Computer Assisted Orthopedic Surgery (CAOS) procedures.
One of the most fundamental tasks in CAOS is the segmentation of bone and cartilage structures in two and three-dimensions. Ideally, automatic or semi-automatic algorithms should operate in a low radiation range, which poses minimal risk to both doctors and patients during imaging. Reduced exposure time correlates with a low signal to noise ratio (SNR), which makes it difficult to interpret images by human eye examination. Image processing and computer vision algorithms should thus be aimed at operating under low SNR in order to highlight, segment and reconstruct relevant structures.
Complexities in the geometric arrangement of bones can pose difficulties for reconstruction purposes. Specifically, imaging of the knee region contains multiple structures with a partially overlapping geometrical arrangement that renders segmentation difficult. To address this problem, and in fact similar problems involving segmentation of multiple rigid bodies, a top-down approach is called for.
To accurately segment and reconstruct the 3D organization of a rigid structure, a volumetric information must first be obtained by modalities such as magnetic resonance imaging. Following steps of registration, which will not be covered here, all bony structures must be accurately extracted. Pre-designed flexible 3D statistical shape models are fitted to the volumetric data by means of optimization methods and their transformation parameters recorded. To assist with fitting, features from the volumetric data can be extracted and used for more accurate estimation of the rigid body transformation parameters. At this point, the placement overlying models on the volumetric data might result in erroneous positioning, which needs to be refined by other methods.
Boundaries of each bony structure can be extracted from cross-sectional information in the volumetric data. The edge are used to refine the positioning and increase accuracy of 3D shapes by utilizing algorithms such as the graph-cut. Following refinement, the ensemble of segmented objects need to be jointly viewed to assess overlaps between close object. Utilization of energy minimization functional to separate overlapping structures should then be performed while taking into account restriction of curvature and deformation of the statistical shape priors.
Correct assignment of energy functional, both for position refinement (separating overlapping objects) and for the graph cuts, is by itself a work of art and it should be subject to the highest rigor, in order to adhere to clinical standards of accuracy. To incorporate algorithmic means for navigation and procedure planning, one must first make sure that segmentation and reconstruction of multiple structures in orthopedic imaging is perfectly accurate. Three-dimensional virtual bone models have great advantages, both in planning a procedure and in the clinical room. The stiff and roughly predictable nature of bones makes them ideal for tracking and for reconstructing a model which can be continuously viewed on screen with instrument overlap (tracked with IR or magnetic means). Such models can perform as the “eyes” of clinicians while carrying out a closed and minimally invasive orthopedic operation.