Summary Statement

Subject specific FE models of human Achilles tendon were developed and optimum material properties were found. Stress concentration occurred at the midsection but dependent on stiffening and thinning of tendon, indicating that they are two major factors for tendon rupture.

Aseptic loosening is the leading cause for revision in total hip arthroplasty. Retro-acetabular lysis is often a silent process until severe bone loss causes catastrophic failure. This presents a technically difficult problem for the surgeon and a poorer result for the patient compared to primary arthroplasty. While the major cause of osteolysis is reaction to polyethylene particles, there is little data on the initiation and progression of such lesions. Further, alterations in the mechanical environment caused by such pathology is unclear. We present our use of 3D, finite element (FE) models of retro-acetabular pathology to investigate the biomechanical effects of osteolysis in total hip arthroplasty. Axial CT scan slices from a patient with cystic osteolysis were selected. Areas of cortical bone, cancellous bone, the cup and the cyst are accurately identified. The axial images are matched to a predetermined grid and used to build a complex finte element model. In this way complex anatomy can be built into the FE model and used to map cystic lesions. Force is then applied to the acetabulum.

Initial analysis shows similar stress transmission in cystic disease compared to the post operative pelvis. Pelvic bone still behaves as a sandwich construct with transmission from the acetabulum to the SI joints, pubic symphysis and medial wall. In the setting of pelvic medial wall deficiency, stress transmission is altered with areas of low stress around the defect.

The FE models containing pathology can be compared to models with generic bone density values immediately after total hip arthroplasty. The presence of a cyst in cancellous bone with intact cortical bone, demonstrates strain patterns similar to the post operative pelvis. Once cortical bone loss occurs strain patterns begin to change. This may mark a critical point in osteolytic progression. We present a developing new tool to be used in the assessment of a patient population with retroacetabular cystic disease.

Patient- specific orthopaedic models are currently used in computer navigation. They provide realistic 3-D geometries for assessment of device placement (e.g. tibial trays, hip implants). Models are generated at time of operation by the surgeon. But patient-specific models have other uses. We envisage a future in which realistic 3-D patient models are routinely used for predicting the outcome of surgical procedures and new devices and for general patient health monitoring.

We are currently developing accurate 3-D models directly from CT scan post-operation. They are being used in investigations of the progress of bone remodeling. Such work can provide valuable feedback on the outcome of new procedures and how bone remodels under load. Such models would eventually include other tissue such as muscles and skin.

But there are a number of research and development challenges associated with the creation of patient-specific models. They include

minimal use of radiation for data collection;

We are working to meet these challenges. They include the use of generic data to supplement patient data, efficient ways of morphing models to fit the patient, and multi-scale modeling strategies. Work in progress at the Auckland Bio-engineering Institute will be presented in this talk.

A number of densitometry studies have reported dramatic density losses in the acetabular region after uncemented Total Hip Arthroplasty (THA)1,2. However the mechanical implication of such loss is not yet known. This study aims to perform a mechanical analysis with patient specific Finite Element (FE) models to find out how the stress distribution affects the Bone Mineral Density (BMD) changes after uncemented THA.

An existing patient CT dataset collected for a densitometry study was used to generate patient-specific FE models with a previously validated FE mesh generation method3. Boundary and loading conditions included the hip joint force and the forces of 21 muscles attached to the pelvic bone at eight characteristic phases of a gait cycle 4. Tensile and compressive components of principal stresses were calculated after each simulation.

In general, both compressive and tensile principal stresses decreased after uncemented THA but the magnitude of decrease for tensile stresses was much greater than compressive stresses. The changes in tensile stresses were matched with BMD loss patterns. In particular, the densitometry study revealed that areas dorsal to the prosthesis lost more bone density than areas ventral to the prosthesis1. The stress distribution pattern showed that such areas experienced high tensile stress initially and then a dramatic decrease in their magnitude while their compressive stresses remained relatively unchanged. On the other hand, the regions where BMD was maintained - the areas superior to the cup - experienced high compressive stresses initially, which remained relatively high three years after the surgery.

Although it is a result from one patient, results suggest that changes to tensile and compressive stresses might influence BMD differently after uncemented THA. Our hypothesis is that regions with high tensile stress experience bone loss while BMD of the regions with high compressive stress are maintained. More patient datasets are being processed to test this hypothesis. Findings from this study can explain the phenomena of retroacetabular osteolysis, late migration and implant failure of press-fit cups observed in long-term clinical studies.

Long term clinical follow-up of total hip arthroplasty (THA) has identified problems associated with cyst formation. Such cysts are formed as a result of expansile osteolysis, which starts on a small area of the skeleton and spreads into the bone away from the surface of the prosthesis. Since large areas of the prosthesis are still in immediate contact with the skeleton the prosthesis is not loose and the patients are usually without pain. However this form of osteolysis may destroy large areas of the skeleton before it is detected and result in a sudden fracture due to a weakened skeleton. While there are some short term prospective trials that have shown changes in bone density in the periacetabular region, one needs a biomechanical model to understand factors that influence bone remodeling leading to cyst formation. This study aims to develop a mathematical model for studying the mechanical effects of bone cysts in the acetabulum of THA patients.

2D finite element (FE) models of patients with known restroacetabular cystic disease were generated using coronal CT images from the central region of the acetabulum. The boundary between bone and soft tissue was segmented and an FE model generated. Mesh convergence tests were performed to identify a suitable level of mesh refinement. Three material zones representing– cortical bone (E=17GPa), cancellous bone (E=1GPa) and a titanium cup (E=120GPa) – were included in the model. A series of simulations were run to investigate how cysts affect stress distribution as well as the mechanical consequence of medial wall deficiency.

The presence of a cyst did not alter the pattern of stress distribution in the lateral and medial wall. But the strain energy function increased significantly at the inferior margin of the cyst within its cancellous bone. This may encourage bone formation at the cyst margin and help to explain the sclerotic walls seen in some cysts. Models with absent medial walls showed that both compressive and tensile stresses lowered in the cortical wall and the strain energy function reduced almost to zero. This suggests that a medial wall defect has a high risk of progression.

The current 2D model cannot incorporate complex acetabular geometry or complex forces acting on the hip. Therefore the current model will be further developed into a 3D FE model of the whole pelvis that also represents the pelvic ring structure more adequately. Physiologically meaningful boundary conditions as well as patient specific geometry and material properties will be used to investigate mechanical effects of bone cysts realistically.

In recent years, some attempts have been made to develop a method that generates finite element (FE) models of the femur and pelvis using CT. However, due to the complex bone geometry, most of these methods require an excessive amount of CT radiation dosage. Here we describe a method for generating accurate patient-specific FE models of the total hip using a small number of CT scans in order to reduce radiation exposure.

A previously reported method for autogenerating patient-specific FE models of the femur was extended to include the pelvis. CT osteodensitometry was performed on 3 patients who had hip replacement surgery and patient-specific FE models of the total hip were generated. The pelvis was generated with a new technique that incorporated a mesh morphing method called ‘host mesh fitting’. It used an existing generic mesh and then morphed it to reflect the patient specific geometry. This can be used to morph the whole pelvis, but our patient dataset was limited to the acetabulum. An algorithm was developed that automated all the procedures involved in the fitting process.

Average error between the fitted mesh and patient specific data sets for the femur was less than 1mm. The error for the pelvis was about 2.5mm. This was when a total 18 CT scans with 10mm gap were used – 12 of the femur, and 6 of the pelvis. There was no element distortion and a smooth element surface was achieved.

Previously, we reported a new method for automatically generating a FE model of the femur with as few CT scans as possible. Here we describe a technique that customizes a generic pelvis mesh to patient-specific data sets. Thus we have developed a novel hybrid technique which can generate an accurate FE model of the total hip using significantly less CT scans.

An automated method of generating FE models for the total hip with reduced CT radiation exposure will be a valuable clinical tool for surgeons.

In recent years, some attempts have been made to develop a method that generates finite element (FE) models of the femur and pelvis using CT. However, due to the complex bone geometry, most of these methods require an excessive amount of CT radiation dosage. Here we describe a method for generating accurate patient-specific FE models of the total hip using a small number of CT scans in order to reduce radiation exposure.

A previously reported method for autogenerating patient-specific FE models of the femur was extended to include the pelvis. CT osteodensitometry was performed on 3 patients who had hip replacement surgery and patient-specific FE models of the total hip were generated. The pelvis was generated with a new technique that incorporated a mesh morphing method called ‘host mesh fitting’. It used an existing generic mesh and then morphed it to reflect the patient specific geometry. This can be used to morph the whole pelvis, but our patient dataset was limited to the acetabulum. An algorithm was developed that automated all the procedures involved in the fitting process.

Average error between the fitted mesh and patient specific data sets for the femur was less than 1mm. The error for the pelvis was about 2.5mm. This was when a total 18 CT scans with 10mm gap were used – 12 of the femur, and 6 of the pelvis. There was no element distortion and a smooth element surface was achieved.

Previously, we reported a new method for automatically generating a FE model of the femur with as few CT scans as possible. Here we describe a technique that customizes a generic pelvis mesh to patient-specific data sets. Thus we have developed a novel hybrid technique which can generate an accurate FE model of the total hip using significantly less CT scans.

Introduction and Aims: CT is one of the most versatile and useful medical imaging modalities for computer assisted surgery (CAS) and monitoring bone remodelling. However, the high radiation dosage hinders its widespread use. We describe a method for generating smooth and accurate Finite Element (FE) meshes using CT data with reduced radiation exposure.

Method: We have performed serial CT assisted osteodensitometry measurement on seven patients who had a total hip replacement. FE models were generated automatically with cubic Hermite basis functions for both geometry and density. The meshes were fitted to the geometric and density data sets using least square’s fitting. Density was displayed over the surface of the elements using a colour spectrum. The effect of reducing radiation dosage was studied by generating five different types of FE meshes from each patient with different numbers of CT slices. The different mesh types were generated by varying the gap between slices.

Results: The mesh with the smallest number of CT slices used seven CT scans, with the gap between slices of 3cm on average while the mesh with the largest number of slices used 22 scans with the gap of 0.8cm. For the mesh with the largest number of CT slices, the average error after the geometric fitting was less than 0.5mm. The average error for the density fitting was 70.2 mg/ml. When expressed as the percentage to the overall density data range (0 ~ 1500 mg/ml), the average error was 4.7%. Meshes generated with a smaller number of CT slices had larger errors, and this increased as the number of slices used decreased. The error in geometry dropped dramatically (more than 50%) when more than 10 slices were used, whereas the error in density decreased approximately linearly as the number of slices increased. Overall, it was possible to generate realistic and smooth meshes with a geometrical error of less than 1.5mm and a density error less than 7% using 10 CT slices.

Conclusion: One strength of the current study is that we have used cubic Hermite elements, which requires much less information in generating FE meshes without sacrificing too much accuracy. Our study has shown that we can generate realistic and smooth meshes with about 10 CT slices of the proximal femur. This is important to enhance the power of CT in clinical applications.

Periprosthetic bone density (BD) changes can be tracked using computed-tomography (CT) assisted osteodensitometry. Patient-specific computer-generated models allow for good visualisation of density changes in bone. We describe techniques for generating smooth and realistic finite element (FE) models that contain both BD and geometry from quantitative CT data using cubic Hermite elements.

FE models were created for three patients who had a total hip replacement. CT-scans were performed at 10 days, one year, and 3 years after the operation and calibrated using a synthetic hydroxyapatite phantom. FE models of the proximal femur were automatically generated from the CT data. Each model had on average 300 tri-cubic Hermite elements. Models were least squares fitted to the entire dataset. BD data was also sampled and fitted using the same cubic interpolation functions. Density was displayed using a colour spectrum.

Realistic patient-specific FE models were obtained. Density and changes in BD were easy to identify. The error in the geometric fitting (RMS distance between data points and the model surface) was generally less then 0.5 mm. The average error for the density fitting (RMS difference between each density data point and the interpolation function value at the same point) was 61.64 mg/ml or 3.08%.

CT osteodensitometry’s potential use as a clinical tool for monitoring changes to BD can be significantly enhanced when used in conjunction with realistic patient-specific finite element (FE) models. Realistic models can be generated with an economic use of scan data, thus keeping radiation dosage down.