A Biofeedback System for Continuous Monitoring of Bone Healing

M. Windolf

, M. Ernst

, R. Schwyn

, S. M. Perren

, H. Mathis

, M. Wilke

and R. G. Richards

AO Research Institute Davos, Davos Platz, Switzerland

Institute for Communication Systems, HSR Rapperswil, Rapperswil, Switzerland

Keywords: Bone Healing, Fracture Healing, Biofeedback, Telemetry, Fracture Monitoring, Smart Implants,

Instrumented Plate.

Abstract: A telemetric biofeedback concept for continuous monitoring of bone healing is introduced. The system is

based on an implantable electronic unit with on-board data processing of deformations or displacements of

fracture fixation devices. In contrast to existing solutions, it allows for autonomous long-term data

collection over several months. The system enables observing fracture motion and patient activity under

daily routine conditions. Feasibility of the approach was proven in an animal experiment with an

instrumented plate monitoring axial motion in a transverse osteotomy gap at the sheep tibia. Callus

formation and maturation of the repair tissue was indicated by a decline of the interfragmentary motion

signal over time and by changes in the animal's activity pattern. For improved understanding and

interpretation of such information, extended collection of in-vivo data is the consequent next step.

1 INTRODUCTION

Flexible internal fixation is an essential modality in

today's fracture treatment. It promotes secondary

bone healing by imposing confined mechanical

stimuli at the fracture site, while still permitting

early recovery of limb function. Having reached a

high level of competence, quick, safe and reliable

healing is achieved in the majority of cases. Despite,

healing disturbances under difficult mechanical or

biological conditions such as infections, non-unions

or osteoporosis remain challenging.

Acceleration of bone healing through mechanical

stimulation of the repair tissue has been investigated

over decades (Claes et al., 1995; Goodship and

Kenwright, 1985; Perren,1979). To define the frame

for an appropriate healing environment, numerous

experimental studies investigated active and passive

mechanical stimulation with varying motion

magnitudes (Claes et al., 1995), motion frequencies

(Hente et al., 2001) and directions of motion (Bishop

et al., 2006; Claes et al., 2008; Epari et al., 2007).

However, in clinical practice the actual mechanical

circumstances at the fracture site remain a black box.

Factors such as individual limb loading, patient

activity, fracture patterns or configuration of the

fixation hardware form a complex setting

influencing the healing outcome. An objective

measure to assess fracture healing under in-vivo

conditions is, hence, required. Such information

could be valuable to improve implant designs and

application to better serve the individual

requirements. Moreover, they could have a

considerable impact on patient care, not only to

accelerate healing, but also to steer weight bearing

and early patient mobilization, detect and react on

healing disturbances or define the appropriate time-

point for implant removal.

Clinical methods to determine the state of

healing are based on radiographic evaluation or

clinical examination. Both are highly subjective.

McClelland et al. (2007) showed considerable

intraobserver variability and overall poor prediction

performance of radiographic healing assessment

criteria. Only weak correlations were found between

the radiographically determined diameter of

mineralized callus and fracture stiffness (Eastaugh-

Waring et al., 2009).

Hence, it was suggested that measuring the load

carried by a bridging implant would be an indirect

but valid criterion to assess mechanical stability of

the fractured bone. It is assumed that the load borne

by the fixation device decreases with ongoing

calcification and stiffening of the fracture callus

while interfragmentary motion and strain within the

fracture gap diminish (Cunningham et al., 1987).

243

Windolf M., Ernst M., Schwyn R., M. Perren S., Mathis H., Wilke M. and Richards R..

A Biofeedback System for Continuous Monitoring of Bone Healing.

DOI: 10.5220/0004913002430248

In Proceedings of the International Conference on Biomedical Electronics and Devices (BIODEVICES-2014), pages 243-248

ISBN: 978-989-758-013-0

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

This load-sharing relation between implant and bone

serves as basis to track the course of healing by

measuring load transmission through implants.

Current telemetric solutions transmitting such

data from inside the body rely on electric induction

as power source. Some are meant as research tools

to measure internal body loads (Bergmann et al.,

2001; Wilson et al., 2009), some target clinical

application (Seide et al., 2012). An induction coil,

positioned at the injured limb, is required for data

collection and transfer and therefore allows only

short term "snapshot" measurements when long term

data acquisition is actually needed for monitoring

the healing progress. Short term measurements of

implant deformation can be disturbed by several

influencing factors such as current physiological

loading of the bone or artificial conditions in a gait-

lab. Natural patient behavior and individual long-

term activity cannot be captured.

An alternative approach is proposed offering

continuous long-term measurements (approx. 4

months) of biomechanical parameters in-vivo

without external power source. An implantable and

autonomously working electronic unit was

developed for continuous recording of fracture

motion under unimpaired natural locomotion. In a

first instance the device was developed as a research

system to analyze the bone healing progress under

selected defined conditions.

2 DATA ACQUISITION CONCEPT

Targeting a method for autonomous long-term data

acquisition, a novel telemetric data logger unit was

developed. The system comprises a microprocessor

for real-time processing of sensor raw data. The

sensor signal is scanned for peak values employing a

custom-made peak detection algorithm. Assuming a

maximum step frequency of 3 Hz, the sampling rate

is set with a tenfold oversampling at 30 Hz.

Together with a stroke counter, peak values are

continuously cumulated. Results are stored

internally at predefined logging intervals. The

influence of natural variances of functional loading

is thereby averaged out. Additionally, the first

derivative of the raw signal is computed in real-time

and the average deformation rate is calculated,

considered as important parameter characterizing

bone healing. To obtain a histogram of loading

intensities, three peak detectors run in parallel at

different amplitude thresholds to sort load strokes

into distinct bins according to their magnitudes.

Instead of storing and transferring the complete

sensor signal, the data is transformed on-board into

small packages of statistically meaningful

parameters (e.g. average amplitude per stroke within

a defined time interval), using an ultra-low power

microcontroller (MSP430AFE253, Texas

Instruments, Dallas, Texas, USA). This approach

follows the hypothesis that the lion's share of such

raw data lacks meaning and would anyway be

discarded at post-processing.

Figure 1: Electronic unit used for on-board processing of

the sensor signal with wireless interface for data tranfer

via radio frequency identification (RFID).

This lean data management allows the use of an

energy-efficient wireless data transfer technology.

The download of the calculated parameters and

settings adjustment (if required) is realized by means

of Radio Frequency Identification (RFID) with a low

frequency transponder (134.2 kHz). Data download

is independent from the data collection process and

can be done on demand at freely chosen time points.

In the current system version, the patient skin is

approached with an RFID transponder to a

maximum distance of 3 cm to the implanted data

logger. The download process for 1 month of

collected data requires 12 min (at 6 h logging

intervals).

Current consumption of the device is ~60 µA

resulting in a battery lifetime of around 4.5 months

(3 V button cell battery with a capacity of

210 mAh). Size of data logger and battery is 26 mm

diameter x 7.5 mm (Figure 1).

The data collection principle is independent of

the processed signal type. Two versions of the

device with adapted signal conditioning have been

realized for receiving signals of different sensors. 1)

Connecting a conventional strain-gauge rosette,

measuring implant/fixator deformation, and 2)

Attaching a miniature LVDT displacement

transducer (linear variable differential transformer)

for measuring fracture gap motion.

In a first application, the LVDT version was used

together with a research implant system in a sheep

tibia model as described in the following.

BIODEVICES2014-InternationalConferenceonBiomedicalElectronicsandDevices

244

3 PILOT ANIMAL STUDY

Functionality of the developed system under in-vivo

conditions was investigated in an animal model.

Purpose of the experiments was proving the

principle, revealing technical and methodological

issues and using the system to answer specific

research questions. Until now, a total of 10 sheep

were operated and equipped with different versions

of the data logger as part of an iterative development

process. Since settings and scope varied between

experiments, a single case will be described in the

following to illustrate the system function. Statistical

evidence is, hence, not presented.

3.1 Materials and Methods

In-vivo measurements were performed in a sheep

tibia osteotomy model. For stabilization of the

fracture a dynamizable internal fixator system was

used (AO Research Institute Davos). The implant is

axially compressible and comprises a proximal and

distal plate-body connected by two cylindrical rods,

which act as linear guides for implant motion. Two

polymer springs (Polyurethane) allow for passive

dynamization of the fracture through weight bearing

and muscle contraction and provide defined load

sharing between bone and implant. Range of motion

can be freely adjusted from rigid blocking to

macroscopic axial displacement. A miniature

displacement transducer (GHSM-1.0B, Singer

Instruments & Control Ltd., Tirat Carmel, Israel,

2 mm measuring range) was incorporated into one of

the guiding rods to measure axial plate motion

(Figure 2). The electronic unit was connected via a

biocompatible cable and was encapsulated in a

custom-made PEEK housing (Polyetheretherketon)

(Figure 3).

Figure 2: Exploded view of the internal fixator

instrumented with a displacement transducer and polymer

springs for passive dynamization of the fracture gap.

Implant and sensor were calibrated on a material

testing machine. Axial motion was limited to

0.3 mm and the springs were preloaded with 250 N.

Axial stiffness of the implant was ~530 N/mm.

Animal experiments were approved by the local

ethic committee for animal health and were carried

out on Swiss white alpine sheep. Surgery was

performed under general anesthesia. The implant

was placed on the medial aspect of the left tibia with

the sensor wire exiting proximally to the electronic

unit, which was positioned in a subcutaneous pocket

proximal to the implant. Following fixation of the

plate with eight 5.0 mm angular stable locking

screws, a 3 mm transverse osteotomy was created

using an oscillating saw and a guiding jig.

Figure 3: Implantation of the research implant system with

telemetric unit in a sheep tibia model. The electronic is

encapsulated in a PEEK housing and inserted proximally

to the axially compressible plate.

The sheep (body weight 50 kg) was able to bear

full weight immediately after surgery. A loose

harness was installed during the first five days

allowing the animal to rest while protecting the

fixation from excessive loading. Radiographs were

taken biweekly until euthanasia at 18 weeks post-

operation.

Thresholds for the peak detectors were set to

motion amplitudes of 0.2 mm, 0.1 mm and 0.04 mm

(0.02 mm after 6 weeks). The logging interval over

which measurements were averaged, was set to 12

hours, thereby generating two sets of parameters per

day (daytime: 6-18 h, nighttime: 18-6 h). The

recorded data was downloaded to an external

computer once or twice a week. After the animal

was killed at 18 weeks post-operation, both tibiae

were harvested and the implant was removed from

the bone. Torsional stiffness and ultimate torque was

determined for operated and contralateral tibia by

means of torsional mechanical testing to failure.

Mechanical behavior of the implant was reevaluated

after explantation and compared to the initial

mechanical test results.

ABiofeedbackSystemforContinuousMonitoringofBoneHealing

245

Figure 4: Recorded axial motion per load cycle over time. Displayed values are averaged over an interval of 12 hours.

Below the corresponding radiographs (AP direction) are shown on the same time scale.

3.2 Results

Data logger, sensor and implant kept functioning

throughout the experiment (18 weeks). No

pathological reactions of the animal to the system

were observed.

Mean axial displacement per stroke increased

from initially 0.1 mm after operation to around

0.2 mm within the first three days. It then

temporarily decreased again to 0.15 mm at day 6,

before reaching the highest displacement of

0.22 mm at day 12. From the beginning of the third

week post-surgery, a rapid reduction of the mean

stroke amplitude was found, followed by a slower

decrease until the minimal threshold of 0.02 mm was

reached after eight weeks post-operation. A step in

the displacement curve after day 40 is caused by

changing the lowest detector threshold to 0.02 mm.

(Figure 4).

A maximum of 480 strokes per hour was

recorded at the second day after surgery (Figure 5).

From day 3 the number of strokes/h decreased and

stabilized at a level of 50 - 100 strokes/h for the

following three weeks before the detected activity

faded out. Number of strokes was consistently

higher during daytime than during night (Figure 5).

While initially the strokes divided equally into

the three intensity bins according to displacement

magnitude, the number of strokes in the highest bin

(≥0.2 mm) increased to 50% within the first 10 days.

During the following two weeks, number of strokes

in the lowest bin (amplitude 0.04-0.1 mm) increased

continuously until no more strokes above 0.2 mm

and above 0.1 mm were recorded after 4 and 5

weeks respectively.

First signs of callus formation on radiographs

were found four weeks post-operation; bridging was

observed after six weeks. Mineralization and size of

callus then gradually increased reaching a maximum

at 8 to 10 weeks post operation.

Torsional stiffness of the operated limb was

4.3 Nm/deg (77% of contralateral). Ultimate torque

to failure yielded 58.5 Nm (72% of contralateral).

At the end of the test axial gliding of the plate

remained possible, but implant mechanics were

found altered by the biological environment. Spring

preload was reduced while stiffness of the plate had

slightly increased.

4 DISCUSSION AND OUTLOOK

A telemetric implantable data collection system for

continuous monitoring of bone healing was

introduced and successfully tested in an animal

experiment. To the authors' knowledge, this is the

first time such data could be acquired over a

complete fracture healing cycle. The general

principle of indirect healing assessment by

measuring fixator deformations (Evans et al., 1988)

was confirmed by a decline of the motion signal

while fracture callus forms. A stable response of the

derived signal over days and weeks supports the

BIODEVICES2014-InternationalConferenceonBiomedicalElectronicsandDevices

246

Figure 5: Number of strokes per hour recorded in time intervals of 12 hours. Green bars represent strokes with an amplitude

less than 0.1 mm, yellow bars strokes from 0.1 mm to 0.2 mm and strokes with an amplitude greater than 0.2 mm are

marked in red. Bars for daytime activity point upwards, nighttime downwards.

underlying assumption that variations in functional

loading can be averaged out. Here 12 h averaging

intervals were chosen; other timespans could be

considered. In the present setting, parameters

derived from the biofeedback system indicated

changes in fracture motion earlier than healing

became apparent on radiographs. Whether this is a

valid observation remains to be clarified. It could

also be attributed to the mechanical behavior of the

research implant used. Preloading the springs acts as

a filter shielding the fracture from low intensity

strokes. The plate is, hence, more responsive to

stiffness changes.

In contrast to other solutions (Seide et al., 2012)

the data logger system also enables assessment of

the patient activity. Onset of bone healing is

reflected by an increasingly unbalanced stroke-

intensity distribution. With ongoing fracture

consolidation, occurrence of high-intensity strokes is

fading out in favor of lower stroke intensities.

Overall patient activity in terms of number of

strokes per time-interval is another interesting

parameter which may contribute to a better

understanding of healing processes in the future. A

pronounced activity of the animal directly after

surgery became obvious, but was not necessarily to

be expected. This cannot be explained with

individual animal behavior since a general tendency

was seen in all tested animals. Pain medication

during the first days after surgery may be a reason

for accentuated limb loading. It is unclear how this

initial activity peak influences the healing process

and if the behavior can also be found in humans.

To better understand such information and to

reach an evidence level, the consequent next step is

to build up a database. Distinct healing patterns for

normal and aberrant courses of healing could be

extracted and interpreted (Burny et al., 2012; Claes

et al., 2002). Ideally, data collection should

concentrate on human patients to increase the

significance of results. Therefore, a strategy is

followed where data can be acquired at minimal

patient risk. As a first step, measurement in

combination with external fixation is an interesting

approach (Claes et al., 2002). A non-implantable,

external prototype has been developed for this

purpose. The device can be attached to an external

fixator measuring sidebar deflections without body

contact. A study to collect clinical pilot data is in

preparation.

Technically, the described data logger is still in a

prototype stage to be used for research purposes.

Several minor issues crystalized during animal

testing including robustness of the firmware, timing

accuracy of the internal clock, cable breakage or

sealing of the housing. The next generation data

logger is currently under development targeting

accelerated data transfer, increased communication

range and miniaturization. A size reduction of the

electronics unit is an interesting future option for a

potential internal application in human patients.

However, this may only be considered at a later

stage, when a distinct analysis of acquired pilot data

has been performed regarding potentials and benefits

for clinical fracture treatment.

Another future concept is the idea of active

modulation of fixation stiffness according to

individual healing progressions. The potential of

such an approach is still unknown; further research

is required (Epari et al., 2013). The electronic unit is

ABiofeedbackSystemforContinuousMonitoringofBoneHealing

247

already prepared not only to connect sensors, but

also actuators to perform defined actions for a

closed-loop control strategy.

5 CONCLUSIONS

A telemetric biofeedback concept for continuous

monitoring of bone healing was introduced. The

system allows for autonomous long-term data

collection over several months, carrying a high

potential to significantly improve fracture care.

Feasibility of the approach was proven in an animal

experiment. Collection of further in-vivo data - also

in humans - is the consequent next step.

ACKNOWLEDGEMENTS

The authors would like to thank N. Ramagnano for

his valuable contribution on electronics layout and

development.

This work was performed with the assistance of the

AO Foundation via the AOTRAUMA Network

(Grant No.: AR2012_02).

REFERENCES

Bergmann, G., Deuretzbacher, G., Heller, M., Graichen,

F., Rohlmann, A., Strauss, J., & Duda, G.N. 2001. Hip

contact forces and gait patterns from routine activities.

J.Biomech., 34, (7) 859-871 available from:

PM:11410170.

Bishop, N. E., van, R. M., Tami, I., Corveleijn, R.,

Schneider, E., & Ito, K. 2006. Shear does not

necessarily inhibit bone healing. Clin.Orthop.Relat

Res., 443, 307-314 available from: PM:16462456.

Burny, F., Burny, W., Donkerwolcke, M., & Behrens, M.

2012. Effect of callus development on the deformation

of external fixation frames. Int.Orthop., 36, (12) 2577-

2580 available from: PM:23073925.

Claes, L., Augat, P., Schorlemmer, S., Konrads, C.,

Ignatius, A., & Ehrnthaller, C. 2008. Temporary

distraction and compression of a diaphyseal osteotomy

accelerates bone healing. J.Orthop.Res., 26, (6) 772-

777 available from: PM:18240329.

Claes, L., Grass, R., Schmickal, T., Kisse, B., Eggers, C.,

Gerngross, H., Mutschler, W., Arand, M.,

Wintermeyer, T., & Wentzensen, A. 2002. Monitoring

and healing analysis of 100 tibial shaft fractures.

Langenbecks Arch.Surg., 387, (3-4) 146-152 available

from: PM:12172859.

Claes, L. E., Wilke, H. J., Augat, P., Rubenacker, S., &

Margevicius, K.J. 1995. Effect of dynamization on gap

healing of diaphyseal fractures under external fixation.

Clin.Biomech.(Bristol., Avon.), 10, (5) 227-234

available from: PM:11415558.

Cunningham, J. L., Evans, M., Harris, J. D., & Kenwright,

J. 1987. The measurement of stiffness of fractures

treated with external fixation. Eng Med., 16, (4) 229-

232 available from: PM:3691938.

Eastaugh-Waring, S. J., Joslin, C. C., Hardy, J. R., &

Cunningham, J. L. 2009. Quantification of fracture

healing from radiographs using the maximum callus

index. Clin.Orthop.Relat Res., 467, (8) 1986-1991

available from: PM:19283438.

Epari, D. R., Kassi, J. P., Schell, H., & Duda, G. N. 2007.

Timely fracture-healing requires optimization of axial

fixation stability. J.Bone Joint Surg.Am., 89, (7) 1575-

1585 available from: PM:17606797.

Epari, D. R., Wehner, T., Ignatius, A., Schuetz, M. A., &

Claes, L. E. 2013. A case for optimising fracture

healing through inverse dynamization.

Med.Hypotheses, 81, (2) 225-227 available from:

PM:23688741.

Evans, M., Kenwright, J., & Cunningham, J. L. 1988.

Design and performance of a fracture monitoring

transducer. J.Biomed.Eng, 10, (1) 64-69 available

from: PM:3347037.

Goodship, A. E. & Kenwright, J. 1985. The influence of

induced micromovement upon the healing of

experimental tibial fractures. J.Bone Joint Surg.Br.,

67, (4) 650-655 available from: PM:4030869.

Hente, R., Lechner, J., Fuechtmeier, B., Schlegel, U., &

Perren, S. 2001. Der Einfluss einer zeitlich limitierten

kontrollierten Bewegung auf die Frakturheilung. Hefte

Unfallchirurg (283) 23-24.

McClelland, D., Thomas, P. B., Bancroft, G., &

Moorcraft, C.I. 2007. Fracture healing assessment

comparing stiffness measurements using radiographs.

Clin.Orthop.Relat Res., 457, 214-219 available from:

PM:17159575.

Perren, S. M. 1979. Physical and biological aspects of

fracture healing with special reference to internal

fixation. Clin.Orthop.Relat Res. (138) 175-196

available from: PM:376198.

Seide, K., Aljudaibi, M., Weinrich, N., Kowald, B.,

Jurgens, C., Muller, J., & Faschingbauer, M. 2012.

Telemetric assessment of bone healing with an

instrumented internal fixator: a preliminary study.

J.Bone Joint Surg.Br., 94, (3) 398-404 available from:

PM:22371550.

Wilson, D. J., Morgan, R. L., Hesselden, K. L., Dodd, J.

R., Janna, S. W., & Fagan, M. J. 2009. A single-

channel telemetric intramedullary nail for in vivo

measurement of fracture healing. J.Orthop.Trauma,

23, (10) 702-709 available from: PM:19858978.

BIODEVICES2014-InternationalConferenceonBiomedicalElectronicsandDevices

248