Monitoring the healing of an internally fixated fractured pelvis using vibration analysis methods

Monitoring the healing of long bones has been studied extensively to reduce the period of encumbrance and unnecessary pain for patients suffering from fractured bones. This is more critical for fractures in the pelvis as the patients can be bedridden for long periods of time to allow proper healing to take place. Patients who suffer from unstable pelvic fractures are usually implanted with internal fixations which provide structural rigidity as well as allow sufficient contact between fracture edges for healing to occur. However this does not provide enough strength to support weight-bearing activities. A 12 week post-operative period of immobility is typically enforced on the patient to ensure this healing process is not hindered. This period of immobility could result in muscle atrophy and may cause more health complications for the elderly. At this point in time, there are no proper ways to evaluate the strength of the healing fracture of a pelvis or guidelines to mobilise patients with unstable pelvic fractures. It is therefore highly beneficial to develop a non-invasive method which can be used in-situ to monitor the healing of the pelvis in hopes of allowing patients to undertake earlier weight bearing activities and reduce muscle degradation. Current methods employed to monitor long bone healing are insufficient for applications in the pelvis as the human pelvis has a significantly complicated geometry which demands a different approach. This thesis reviews the different methods which have been used and are currently being used to monitor healing in bone, ultimately proposing the use of vibration analysis techniques as previous studies and vibration theory have shown promising results. Vibration analysis methods are able to assess the strength and integrity of a structure globally, which may be able to provide in-situ monitoring of fracture healing in a human pelvis. By relating the dynamic behaviour of a healing pelvis to its integrity, this method can be incorporated to a quantitative and non-invasive monitoring system. This thesis tests the vibration analysis methods using experimental as well as finite element methods. Experimental tests were conducted on 4th generation synthetic pelvises instrumented with an array of PZT sensors. A constraining rig was developed to support the pelvis such that it accurately represents constrains of the pelvis in-vivo. The synthetic pelvis was cut at the sacrum to simulate a fractured pelvis followed by the application of araldite epoxy to simulate healing by allowing the epoxy to cure. Excitation signals were introduced to the synthetic structure using both impact excitation as well as shaker excitation with the fixation pins as wave guides. Measurements were collected from the sensor array over the curing period to obtain the response functions. A comparison of the response functions against time allowed for the araldite to cure demonstrates significant changes in frequency as well as amplitude of resonant features over the stiffness recovery period. These changes were eventually quantified by calculating stiffness restoration parameters using the response functions, using the response of the pelvis in its cut state as a baseline. The parameters were found to be directly proportional to the glue curing time, where the recovery parameter increases with increasing curing time. This study however, was incomplete as the recovery index was not related to an absolute parameter of strength or integrity of the pelvis. Vibration analysis methods were also initially proposed due to their potential of detecting sub-surface fractures and non-uniform healing. Additional experiments were conducted to test the reliability as well as sensitivity of this method for detecting non-uniform or partial healing in a pelvis. Findings from the secondary experiments were able to conclude that vibration analysis techniques were sensitive enough to detect the partial healing. The recovery parameters calculated from partial healing experiments were almost consistently lower than those calculated from initial tests. To solve the problem of relating these stiffness recovery indices to an absolute parameter of strength however required the use of finite element methods. A pelvis with a Dennis I fracture was modelled with its stiffness and density of bone around the fracture site varied to simulate stiffness restoration which occurs during healing. Up to 50 mode shapes for each healing state were computed to obtain a response within a frequency range comparable to those obtained experimentally. The frequency response functions were extracted from different locations around the pelvis and fixation for analysis and were later compared to results obtained from experiments. The responses seem to reflect dynamic changes which are proportional to the changes in stiffness and density of the bone properties at the fracture site. It was also found that certain resonant features which were driven by the fixation were very prominent in both FE as well as experiments in addition to being sensitive to stiffness changes at the fracture. This finding was very promising as it could allow the monitoring system to interrogate the health of the pelvis without direct attachment to the pelvis. For completion of this research, a final study was done via FE to create a more realistic healing model of the pelvis. This independent study shows results which reflect findings obtained both experimentally and numerically. The dynamic responses obtained from this model were shown to vary proportionally to improving stiffness at the fracture.