Robotic-assisted orthopaedic surgery

There are many surgical procedures that require precise alignment of a tool, such as a drill or needle, based on medical imaging of internal structures. These trajectory dependant processes are found in procedures ranging from orthopaedics (drilling of locking holes for internal fixation of fractures) to neurosurgery (evacuation of subdural haematomas).

This thesis presents an investigation into the automation of trajectory based surgical procedures.

This work is evaluated based on the exemplar procedure of antegrade nailing of femoral shaft fractures, but the individual problems are examined in a more general sense.

In this manner, the solutions posed have been evaluated experimentally for the exemplar procedure while developing solutions that are applicable to a much wider field.

There are many surgical procedures that appear to be ideally suited for robotic assistance or guidance.

The main features of highly suitable procedures include the requirement to align a tool accurately with only internal imaging.

This describes a large number of components of surgical procedures; however adoption has been very slow.

There are barriers to entry that many surgical procedures have in common that have so far resisted automation such as the implementation of robotic assistance; these include system complexity and integration into surgical workflow, retraining requirements, system’s possible negative effect if safety protocols are not adhered to or inadequate, short term cost implications, system datuming and accuracy.

The main results from this research to address, in part, the aforementioned barriers are 1)~a method for the modelling of X-Ray subject environments to determine optimum filters for best contrast, 2)~a fault tolerant, flexible, image registration method, 3)~an accurate but inexpensive method for robot calibration, and 4)~a method for integrating a robotic surgical assistant into existing surgical workflow.

A method for improving the visibility of objects of interest in X-ray images is investigated.

This is achieved by shaping the beam spectrum to the elemental composition of the scene in order to give higher contrast to features of interest relative to the background.

This is accomplished by simulating the absorption by interesting and non-interesting features of the X-ray subject in order to select a filter composition that produces the optimum beam spectrum.

However, the filtering material and thickness have to be appropriately selected for the structure which is imaged. For example, using 0.2mm samarium gave increased contrast/visibility for steel fiducials when viewed through muscle and bone.

Another requirement for robotic assistance is imaging of the operation site for datuming (registration) purposes.

However, there is distortion with X-ray imaging, with older X-ray units cause higher distortion compared to later ones.

There is also the problem of missing datuming fiducials on the image.

Current point matching techniques are unreliable with the combination of distortion and missing/extra points in X-ray images.

Image matching techniques are also unreliable because of small scale/directional features cannot be detected in X-ray images.

Current robotic surgery methods overcome these difficulties by customising the system to suit specific tasks/operations, e.g.\ using special fiducial arrangements to suit the operation.

A fault tolerant, flexible, generic image registration method is proposed to overcome both the effect of distortion and missing markers on the image.

The developed algorithm, named ``Travelling Encampment Algorithm'' provides robust point matching that is tolerant of distortion, missing points and false positives.

This allows registration of X-ray images to be performed with planes of fiduciary markers with great flexibility in the positioning of the X-ray C-arm and X-ray subject without the requirement for special fiducial arrangements.

This technique was compared to existing methods (such as Scale-Invariant Feature Transform [SIFT], Speeded Up Robust Features [SURF] and Iterative Closest Point [ICP], Coherent Point Drift [CPD]) and it worked automatically without human intervention, while others failed or required human intervention to datum the image.

Another need for robotic applications in surgery, or any industrial application requiring precision, is accuracy of the end-effector pose and/or motion.

Modern manufacturing methods make it relatively easy to produce small and very precise robots, but in order to be used without the constant verification of the pose of the end-effector, they must be calibrated accurately.

As robots become smaller and lighter, regular verification and calibration becomes also more important.

There are many calibration techniques/systems which can be used to check the accuracy of a robotic system, but these tend to expensive, large, and requiring special equipment/ environment.

A new calibration method is proposed which is suited to the accurate calibration of small inexpensive robots designed for trajectory based procedures such as drilling and needle insertion. The developed calibration method is based on the optical measurement of a laser beam tracked using off-the-shelf flat-bed scanners.

The technique involves the measurement of the laser spots (on the scanner) with respect to a calibrated frame that is visible as a border in the scans.

This is used to calculate the pose of the end-effector and is especially accurate in determining the trajectory of the end-effector, which is the most important factor in drilling and needle insertions.

A four degree of freedom robot has been used to evaluate the system.

The calibration method has been shown to be capable of measuring the Z axis of the end-effector of a four degree of freedom robot with an accuracy within a cylinder with 0.2mm diameter.

Finally, a potential method for implementing robotic assistance into the workflow of the exemplar procedure is proposed.

This ensures that a change to robotic assistance could be smoothly integrated, as time in surgery is negatively correlated with recovery time.

A four degree of freedom passive autoclavable manipulator, and an automatic X-ray capture, machine vision and trajectory planning system are proposed.

The method developed to extract the nail pose (for intramedullary nailing procedures) is generic; it can also be used for other fixation devices of different shapes and sizes.

This method, which is based on real-time simulation of the nail and X-ray cone, a requires only two X-ray images (one lateral and one Anterior-Posterior), compared to the existing circle matching method, which often requires the radiographer to take multiple lateral images in addition to Anterior-Posterior images.