Generation of subject-specific, dynamic, multisegment ankle and foot models to improve orthotic design: a feasibility study

Inverse dynamic model

Standard gait analysis and inverse dynamic models of the lower extremity consider the foot as a single rigid segment. Some models have been developed that describe the foot in more detail, however to the authors' knowledge no kinematic model has attempted to describe all 26 bones. The proposed kinematic model is scalable and parametric and will integrate into the existing AnyBody whole body musculoskeletal model (Figure 1). The model will contain all of the ligaments and muscles of the foot and ankle. By combing the inverse dynamic modelling with an optimisation algorithm the model will provide insight in function of the foot and leg muscles during gait.

Forward dynamic model

The forward dynamic model is a combined multibody and FE approach. Bones, joints, ligaments and muscles are defined as multibody elements. This multibody model is surrounded by a freeform FE sheet, representing the skin. The soft tissue characteristics are implemented by loading functions that connect the sheet to the multibody elements. This combination leads to a computationally less complex model compared to a full FE model of the foot and ankle. This reduction could be improved by a full multibody representation of the ankle-foot complex, however in this model computation of the plantar pressure would be impossible. In the proposed model this is solved by a FE mesh representing the plantar surface, leading to the possibility of performing a dynamic simulation to compute the deforming plantar surface.

An existing model (Figure 1) has previously been developed from a combination of data from a post-mortem human subject and information derived from existing literature [15]. The generation of one complete dataset leads to a consistent model. Beyond this, the generation of datasets from several subjects creates the possibility of scaling the model. This scaling will be extended for patients requiring foot and ankle orthoses.

The overall aim of this work is to develop detailed and accurate biomechanical models of the foot and ankle which can be used to inform the design of foot and ankle orthoses by predicting the biomechanical effects of the device.

Inclusion criteria

In both groups, healthy feet and group pathological feet, participants will be included if they are physically able to walk at least 20 meters barefoot and unaided. In addition, participants must be in age group 18-50 years and have feet of size 38-44 (EUR). All participants will be fully competent and be able to give informed consent.

In addition to this general inclusion criteria, participants in the patient group must fall into one of the following categories: 1) Patients with metatarsalgia on their right foot, who would be prescribed pressure relieving foot orthoses. Metatarsalgia will be clinically diagnosed as the presence of one or more features of spontaneous pain or tenderness at one or more metatarsophalangeal joints elicited by firm pressure and/or movement of the joint. 2) Patients with flexible flat foot deformities on their right foot, who would be prescribed foot orthoses to improve alignment. For the purposes of this study, flexible flat foot is diagnosed as a correctable relaxed calcaneal stance position greater than six degrees everted with a navicular tuberosity height lower than 35 mm. 3) Hemiplegic stroke patients who would be prescribed an ankle foot orthosis on their right foot to control the motion of their ankle during gait.

Exclusion criteria

Participants in both groups will be excluded if they have a diagnosable disease with known involvement of the lower limb and foot including, for example, diabetes mellitus, peripheral vascular disease and rheumatoid arthritis. Pregnant and lactating women will not be eligable for participation due to the radiation dose associated with the CT scans that are part of this protocol.

To be eligible for inclusion in the healthy group, participants must not currently be receiving treatment for any foot or ankle conditions or have any significant history of foot or ankle trauma, injury, fracture or dislocation.

Kinematic measurements

Kinematic data will be collected using a 12 camera Qualisys system (Glasgow) and an eight camera Vicon system (Maastricht). Residual errors of < 1 mm are deemed acceptable for both systems.

For the shod trials, the participants are provided with standardised footwear (Flextop Diabetic Black shoes, Reed Medical, Blackburn, UK). A number of the foot mounted markers used in the barefoot trials will be removed for the shod trials (see Table 2 for details), and holes are cut into the shoes to allow the remaining markers to be visualised by the motion capture system and to move during walking without interference.

Kinetic measurements

Kinetic measurements will be taken at both centres using Kistler force plates (Kistler Instrument Corp., Amherst, NY) synchronised with and recorded through the QTM software (GCU) or Nexus software (MUMC+) at a frequency of 2400 Hz.

Electromyographic measurements

All parts of the protocol relating to surface EMG measurement will be carried out in accordance with the guidelines produced by the Surface Electromyography for the Non-Invasive Assessment of Muscles (SENIAM) project [20]. These guidelines cover the location and orientation of electrode placement, skin preparation and signal tests for each muscle.

Trigno wireless EMG systems (Delsys Inc, Boston, MA) will be used to collect the EMG measurements at both centres. The electrode units will be attached to the following muscles: tibialis anterior, gastrocnemius medialis, gastrocnemius lateralis, soleus, peroneus longus, vastus lateralis, rectus femoris, and biceps femoris. Signals from each muscle will be checked in real time using EMGworks software (Delsys Inc, Boston, MA) while performing the exercises described in the SENIAM guidelines.

Reference measurements will be taken for each muscle in the form of maximal voluntary isometric contractions (MVICs). Measurements will be recorded for five seconds in total with the participant being asked to gradually build up the force they apply over the first two seconds, and maintain their maximum effort for the remainder of the contraction. Each contraction is repeated three times in a non-consecutive randomised order with at least one minute recovery time between exercises.

For the MVIC and gait components of the testing, EMG signals from the Trigno sensors will be recorded through the analogue channels of QTM or VICON software at a frequency of 2400 Hz.

Plantar pressure measurements

For barefoot trials, plantar pressure measurements will be collected using a 0.5 m Footscan^® plate (RSscan International, Lammerdries, Belgium) recording at 500 Hz. The plate is mounted directly on top of the force plate and secured in place using double sided tape. The effect of this setup on the accuracy of the force plate was assessed using the CalTester^® quality assurance tool (C-Motion Inc, Germantown, MA) and errors were found to be within acceptable limits. To ensure high levels of accuracy in the pressure measurements, the pressure plate is dynamically calibrated with the vertical force signals from the force plate.

In-shoe plantar pressure measurements will be made using the Pedar^® system (Novel, GmbH, Munich, Germany) at 50 Hz. In addition, pressure under the sole of the shoe will be recorded during these trials using the Footscan^® plate.

Testing

A static trial will be recorded with the participant in a relaxed standing pose in the motion capture area. The participant will then be asked to walk barefoot at a self selected speed along the motion capture area such that their right foot strikes the centre of the pressure plate. Five successful walking trials will be recorded. This will then be repeated for the shod trials.

Computed tomography (CT)

CT scans of the leg and foot will be undertaken in hospital radiology centres. Each participant will undergo four scans under a variety of different conditions. First the knee, shank and foot will be scanned with the participant asked to apply only minimal loading to the plate with the foot in a neutral position (90 degrees flexion). Then three scans imaging the foot and ankle only will be taken in a randomised order: foot loaded to 50% body weight with the foot in the neutral position; foot loaded to 50% body weight with the foot 25° of plantar flexion; 50% body weight with 10° of dorsiflexion.

To allow the participant to apply force on the foot while being scanned, a novel loading rig was developed and fabricated (Figure 2). Previously, several investigators have described similar systems (14) but to the authors' knowledge this is the first device that allows the position of the foot in relation to the lower leg to be easily and quickly manipulated.

The loading rig takes the form of a chair fixed on to a plywood base plate and a plate for the participant to push against with their foot at the other end. The nature of CT means that metallic objects can cause interference in the image, an effect known as scattering. To avoid this, the steel components which allow the loading plate to be repositioned are kept behind the scan plane. The loading plate can be easily moved closer to or further from the chair at fixed 20 mm intervals to suit the participant's leg length. A standard bathroom scale with a large LED display is mounted on the loading plate and is used to provide feedback to the participant on the level of loading being applied. The relatively short time frame of the CT scan (5-10 seconds) allows medium level loads to be applied and maintained over its duration. During the scans, a Pedar pressure insole (Novel, GmbH, Munich, Germany) is placed between the foot and the loading plate. In addition, using the pen marks made on the foot to guide placement, radiopaque markers (4 mm diameter Beekley Spots^®, Oncology Imaging Systems, East Hoathly, UK) are attached at the same points as the motion capture markers during gait analysis.

The following parameters were used to acquire the four scans on an Aquilion 64 slice scanner (Toshiba, Tokyo, Japan) at the Glasgow centre or a Brilliance 64 slices (Phillips, Amsterdam, Netherlands) at Maastricht: 120 kVp, 100 mAs, 1.0 mm collimination, 1.0 mm effective slice thickness, pitch factor 41, rotation time 0.5 seconds, B30S medium smooth reconstruction kernel, 512 × 512 matrix.

MR

Due to the limitations of soft tissue information that can be inferred from CT imaging, a subset of five participants will also have MR scans taken of their right foot. This will take place on a different day to the rest of the testing, but within six weeks of the initial assessment.

For the MR scans the foot is placed in a suitable imaging coil and foam padding is placed around the foot to prevent movement during the scan. Images will be acquired using a 3T system (Siemens Verio; Erlangen, Germany). Three scans will be taken in total, a T2-weighted scan covering the full foot and ankle, and two T1-weighted scans, one of the rearfoot/midfoot complex and one of the forefoot/midfoot complex.

Scanning parameters for the T2 scan (trueFISP 3D volume) are: repetition time, 9.8 ms; echo time, 4.92 ms; flip angle, 35°; field of view, 290 mm; slice thickness, 0.6 mm (no slice gap); slices per slab, 144; matrix, 256 × 216 (interpolated); phase encoding, anterior to posterior; number of averages, 2.

Scanning parameters for the T1 scan (Space 3D volume) are: repetition time 700 ms; echo time 22 ms; flip angle 105°; field of view, 150 mm; slice thickness 0.8 mm (no slice gap); slices per slab, 94; matrix 320 × 290 (interpolated); phase encoding anterior to posterior; averages, 2.4.

Gait analysis data

The kinematic data, the kinetic data and the plantar pressure will be processed by using Nexus (Vicon, Oxford, UK) software (Maastricht) or Qualisys Track Manager (Qualisys, Gothenburg, Sweden) software (Glasgow) and saved in the C3D format.

Kinematic and kinetic measurements will be used as an input for the inverse dynamic model and to validate the forward dynamic model. The dataset of one healthy subject will be used to develop the first model. Afterwards the datasets of the other healthy subjects will be used to refine the initial model, develop kinematic and kinetic rules and a morphing-based scaling facility so the models can be personalised. After the development of the morphing algorithm the combination of this algorithm and the two models will be used to predict the effect of insoles.

Data from the foot assessments in combination with the 3D surface scans will be processed to develop an algorithm for personalisation of the models with a non invasive, low-end method. It is intended to investigate if an algorithm can be developed to drive the musculoskeletal model directly from plantar pressure measurements, allowing the kinematic parameters of the lower extremities to be determined without the need for a full motion capture system.

Imaging data

CT data will be segmented into the individual bones of the foot using Mimics (Materialise, Leuven, Belgium) image processing software. The CT-data will be used to compute the joint axes of the foot, by correlating the positions of the bones in the various positions. The loaded CT-data in combination with the pressure measurement will be used to gain insight in the soft tissue characteristics of the foot. Insertion and via points will be identified from partial segmentation of the MR data and described as co-ordinates on the bones segmented from the CT data. Insertion points will be defined as the centre of the area of insertion.