Image-free navigation technology relies heavily on the surgeon carefully registering bony anatomical landmarks, a critical step in achieving accurate registration which affects the entire procedure. Currently this step may depend on placing a pointer superficially, with soft-tissue and skin obscuring these bony landmarks. We report initial results of using newly developed experimental software which automatically recognises the bone soft-tissue interface. This is the first critical step in development of automatic computer generation of the bone surface topography from ultrasound scanning. Individual 2D ultrasound images (n=651) of the anterior femoral condyles and trochlear notch were used. Images were taken from 29 volunteers (20 male, 9 female). The software extracted bone-soft tissue interface by a two-step method based on a gradient evaluation and the elimination of false-positives with a graph closure. The trochlear notch was automatically defined by geometrical modelisation. Coordinates of both bone interface and trochlear notch position for each separate image were compared to a separate analysis performed manually by a single investigator. Error was calculated using root mean squared (RMS). Median error (RMS) in locating bone soft-tissue interface was 0.67 mm, (mean 0.93 mm, SD 0.84 mm). Median error for trochlear notch topography was 1.01mm, (mean 1.41 mm, SD 1.37 mm). Bone soft-tissue interface can be accurately defined and displayed by this software. Direct visualisation of critical bony landmarks could replace the current comparatively subjective placement of a pointer on superficial tissues. This has powerful application in both non-invasive and surgical computer-assisted acquisition of knee kinematics, and may have further applications in orthopaedic surgery.
To study intra- and inter-observer variability with the use of the ultra-sound transducer and percutaneous digitiser point probes To assess the learning curve with the use of the ultrasound transducer probe As part of a larger cadaver study evaluating navigated total hip replacement via the posterior approach, we assessed data relating to acquisition of bony landmarks of the Anterior Pelvic Plane (APP) by four surgeons with an ultrasound transducer and a percutaneous point probe. The surgeons had differing levels of experience with hip surgery in general, and also with surgical navigation per se, but none of them had previously used the ultrasound probe for the specific purpose of landmark acquisition. Without fixing an absolute positional value for any of the bony landmarks, the points registered for individual landmarks by each surgeon were then studied, looking at the three-dimensional spread of these points relative to each other about the mean value. The data from all four surgeons were analysed, looking at the global dispersion of points acquired by the ultrasound and percutaneous point digitiser probes. Our results show that with the exception of a few isolated outliers, the ultrasound probe generated values fell within a +/− 10 mm range. For all four surgeons, the global spread of ultrasound-registered points was noted to be less than that acquired by percutaneous point probe acquisition. Of interest was the finding that points registered by individual surgeons using the ultrasound probe tended to be grouped distinctly together but spatially separate from those of the other surgeons; it would appear that each operator was “homing” in on what he perceived to be the bony landmark in question on the projected ultrasound image. With the percutaneous pointer probe, and with the anterior superior iliac spines as the target, there was closer grouping of points around the mean positional value for the two surgeons who were experienced with its use. However, at the symphysis pubis, the spread of points for these surgeons were not much different from the other two less experienced one, with these points showing a global spread as great as 25 mm. Regardless of the experience of the surgeon, the use of the ultrasound transducer probe appears to be more accurate than percutaneous pointer probe for acquisition of the bony landmarks that constitute the anterior pelvic plane. The learning curve associated with its use is seemingly short and steep. Its accuracy is limited by the fact that the identification of the bony land marks on the on-screen display is open to interpretation by the individual. Methods to standardise the identification of these landmarks on ultrasound images may help improve its accuracy in the future.
