In orthopedic surgeries, it is critical to reduce the risks of drilling complications during bone fracture fixation, especially around critical organs such as in acetabula-pelvic procedures. Either over-drilling or x-ray overuse shall be avoided to reduce potential complications to the surrounding critical organs or tissues. Toward recognising perforation process during bong drilling, we employed drilling vibration signal analysis based on the measurements from miniature inertial sensors. Time-frequency analysis is used for features extractions, which show that information from drilling vibration measurements could reveal the drilling process, hence help doctors track the drilling process and avoid over-drilling.

We addressed the aforementioned challenges through inertial sensor development, vibration measurements, and time-frequency signal analysis. In the preliminary ex-vivo bone drilling experiment setup, an inertial sensor is mounted on a pig femur bone with two fixing nails and can capture 3-axes acceleration data during drilling procedures. A cordless drill is used with Kirschner wires (K-wires) and the diameter of the pin is 3.5 mm. The mounting locations of inertial sensors are close to actual drilling entries without affecting normal procedures. The recorded vibration signals indicate how the drill is interacting with surrounding bone tissues, which shall have different patterns along the deep drilling process. After normalisation, the power spectral density (PSD) is calculated to examine the frequency domain representation of the time series during drilling process. As the drilling vibration process along the bone is non-stationary, we further employ wavelet transform for more localised time-frequency analysis.

When the bone substance interacts with drill bits, compact substance and spongy substance have different bone densities and structures, thus inducing different vibration waveform patterns. In our preliminary experiments, we recorded acceleration data from the pig femur drilling process, where a surgical drill penetrates from compact substance, spongy substance and then to compact substance again. The article shows the feasibility study of estimating femur bone drilling process based on vibrations signals captured from low-cost miniature inertial sensors. Through a preliminary animal ex-vivo bone study, the proposed framework of time-frequency wavelet analysis indicates the drilling interface between compact substance and spongy substance. It shows potentials in perforation recognition along drilling process and more clinical studies will be performed for validating its capability in over-drilling avoidance.

Surgical navigation systems enable surgeons to carry out surgical interventions more accurately and less invasively, by tracking the surgical instruments inside human body with respect to the target anatomy. Currently, optical tracking (OPT) is the gold standard in surgical instrument tracking because of its sub-millimeter accuracy, but is constrained by direct line of sight (LOS) between camera sensors and active or passive markers. Electromagnetic tracking (EMT) is an alternative without the requirement of LOS, but subject to environmental ferromagnetic distortion. An intuitive idea is to integrate respective strengths of them to overcome respective weakness and we aim to develop a tightly-coupled method emphasising the interactive coupled sensor fusion from magnetic and optical tracking data. In order to get real-time position and orientation of surgical instruments in the surgical field, we developed a new tracking system, which is aiming to overcome the constraints of line-of-sight and paired-point interference in surgical environment. The primary contribution of this study is that the LOS and point correspondence problems can be mitigated using the initial measurements of EMT, and in turn the OPT result can provide initial value for non-linear iterative solver of EMT sensing module.

We developed an integrated optical and electromagnetic tracker comprised of custom multiple infrared cameras, optical marker, field generator and sensing coils, because the current commercial optical or magnetic tracker typically consists of unchangeable lower level proprietary hardware and firmware. For the instrument-affixed markers, the relative pose between passive optical markers and magnetic coils is calibrated. The pose of magnetic sensing coils calculated by electromagnetic sensing module, can speed up the extraction of fiducial points and the point correspondences due to the reduced search space. Moreover, the magnetic tracking can compensate the missing information when the optical markers are temporarily occluded. For magnetic sensing subsystem comprised of 3-axis transmitters and 3-axis receiving coils, the objective function for nonlinear pose estimator is given by the summation of the square difference between the measured sensing data and theoretical data from the dipole model. Non-linear optimisation is computational intensive and requires initial pose estimation value. Traditionally, the initial value is calculated by equation-based algorithm, which is sensitive to noise. Instead, we get the initial value from the measurement of optical tracking subsystem. The real-time integrated tracking system was validated to have tracking errors about 0.87mm. The proposed interactive and tightly coupled sensor-fusion of magnetic-optical tracking method is efficient and applicable for both general surgeries as well as intracorporeal surgeries.

Despite of the significance of computed tomography (CT) images in surgery planning and guidance, CT scans are not always applicable due to high radiation exposure, particularly risky for children and youth. It is critical to reduce radiation exposure for high sensitive candidates and statistical atlas based approach has therefore been an alternative with minimal radiation exposure.

We addressed the aforementioned challenges through statistical atlas constructions, 3D atlas to 2D radiography registration to get patient-specific models with minimal radiations and multiple-objective optimisation for planning the treatments. Statistical atlas can be employed to construct the global reference map. The atlas then can be registered to a pair of intra-operative fluoroscopy images for constructing a patient-specific model. In this way, we can reduce the radiation exposure to the patients significantly. To characterise shape variations, a statistical shape atlas is constructed using Point Distribution Model, by which a mean shape, modes of shape variation and shape variation are obtained. To construct the patient specific model from the statistical atlas, 3D-2D registration is essential and a back-projected ray based 3D-2D Iterative Closest Point registration method is investigated. Then the treatment planning module for optimal insertion is investigated to avoid critical zone and unnecessary punctures.

The experiment shows the feasibility of the proposed method for atlas-based, image-guided orthopaedic interventions using minimal radiograph and optimal planning. The proposed framework can be extended to other potential applications and one example is for periacetabular osteotomy, particularly for young females which is of great importance to minimise radiation dose during surgical planning and navigation.