AUTOMATIC SCREENING OF ACCELERATION SIGNAL

DURING PIVOT-SHIFT TEST BASED ON PEARSON’S

CORRELATION COEFFICIENT

Cecilia Signorelli, Nicola Lopomo, Simone Bignozzi, Stefano Zaffagnini and Maurilio Marcacci

Biomechanics Lab, Rizzoli Orthopaedic Institute, 1/10 Di Barbiano Street, Bologna, Italy

Keywords: Pivot-Shift test, ACL reconstruction, Acceleration, Pearson’s correlation coefficient, Template.

Abstract: Anterior cruciate ligament injury produces a pathologic kinematics of the limb that can lead to the evidence

of a pivot-shift (PS) phenomenon. PS-test, specifically performed to highlighted this knee dynamic

instability, is however difficult to quantify. From a clinical point of view is therefore mandatory to find a set

of parameters able to quantitatively characterize PS phenomenon, thus distinguishing between pathologic

and healthy knees. This study proposed a methodology able to automatically quantify PS phenomenon,

analysing the signal recorded by means of a tri-axial accelerometer while executing PS-test itself. A signal

template, which reproduced the 3D acceleration average trend while PS phenomenon occurs, was passed

along the signal in order to recognise the presence of similar patterns. The recognition of the signal

interesting share was based on the calculation of the Pearson’s correlation coefficient between the template

and the corresponding part of the windowed signal. The data acquisition concerning to the first 35 patients

was used to testing the template; in this analysis we considered both the data relative to pathologic and

healthy knee, as well as pre- and post-anaesthesia data, in order to evaluate the influence of active muscular

resistance. The methodology followed had assured a recognition of PS repetitions with an accuracy of

96.7%, a sensitivity of 81.9% and a specificity of 99.3%; therefore can be considered a valid and easily

computable method for the automatic screening of the acceleration signal during PS test. In the future this

method will be uptake in order to quantify the possibility to discern between pathologic and healthy knee.

1 INTRODUCTION

Scientific literature agrees with the fact that pivot-

shift (PS) phenomenon is one of the essential sign of

knee dynamic instability, that can specifically

highlight an injury of the anterior cruciate ligament

(ACL). Moreover clinical results of PS test are more

correlated with functional outcomes, patient's

satisfaction, joint instability and relief of symptoms,

whereas clinical examinations addressing static

instabilities (i.e. Lachman and drawer tests) are less

related to patient’s subjective status (Kocher, et al.,

2009).

The main problem in using PS test is associated to

the difficulty of quantifying the test outcome; this

difficulty is in fact due to the complexity of the

manoeuvre itself, being based on a valgus stress

while flexing the limb and maintaining the foot

internally rotated (Jacob, et al., 1987). In fact,

surgeons who carry out PS test, usually affirm to

sense a sort of 'clunk' while testing a pathologic

knee; this qualitative evaluation is extremely

difficult to quantitatively define and thus

quantitatively differentiate an ACL-injured limb

with respect to a healthy one is a not simple task.

During the last decades qualitative methods and

several devices have been suggested to standardize

the level of ACL injuries correlated to joint dynamic

instabilities, but they revealed a quite complex and

bulky set-up for an outpatient’s clinical use or

allowed only a partial assessment of joint laxity

(Lopomo, et al., Kuroda, et al., Dierman, et al.,

Labbe, et al., 2008; Amis, et al., 2008, 2005;

Mushal, et al., 2007; Csintalan, et al., 2006; Bleday,

et al., 1998).

For these reasons a method which can provide a

reliable diagnosis of dynamic knee instability with

fewer and simpler measurements, can surely be of

benefit for clinical scope.

236

Signorelli C., Lopomo N., Bignozzi S., Zaffagnini S. and Marcacci M. (2010).

AUTOMATIC SCREENING OF ACCELERATION SIGNAL DURING PIVOT-SHIFT TEST BASED ON PEARSON’S CORRELATION COEFFICIENT.

In Proceedings of the Third International Conference on Bio-inspired Systems and Signal Processing, pages 236-239

DOI: 10.5220/0002720402360239

 SciTePress

As reported in literature (Colombet, et al.; Hoshino,

et al., 2007), one of the most interesting parameter to

investigate can be the acceleration reached by the

joint during the PS manoeuvres itself, representing a

possible correlation with the alteration in the fluidity

of the movement due to the presence of PS

phenomenon. The main objective of this study was

therefore to develop and validate a methodology

able to automatically quantify the PS test, giving

also a clinical reliability to the analysis. In particular

3D acceleration was recorded during PS manoeuvres

by means of a commercial triaxial accelerometer and

the output signal was automatically analysed in

order to filter the noisy part and to automatically

extract the clinical information related to the PS test.

2 METHODS

We conducted the presented study on 35 consecutive

patients, with acute or chronic ACL injury, that

underwent ACL reconstruction between August

2008 and August 2009 in our Institute. The

performed experimental procedure was established

on the first 15 patients that were not included in this

study. The applied clinical protocol was approved by

the Institutional Review Board of the Institute.

The system used to acquire 3D joint acceleration

during PS test was a triaxial accelerometer sensor

(Inertis-link, Microstrain Inc., Williston, VT, USA);

the device was connected wirelessly to a dedicated

laptop on which, by means of a simple interface, the

operator was able to monitor in real-time the

patterns of acceleration and to acquire the 3D data.

The sensor was not-invasively mounted on the skin

of the patient and securely fixed by means of a

specific strap between the lateral aspect of anterior

tuberosity and Gerdy’s tubercle, in order to achieve

an optimal stability. The PS test was repeated 3

times by an expert and a novice orthopaedic

surgeon. Each surgeon realized test repetitions, both

on pathologic limb and healthy one. The test was

repeated by the expert surgeon also before and after

the general or regional anaesthesia.

We used the module of 3D acceleration measured by

the sensor, as the most significant parameter to be

analysed during the PS manoeuvre. In particular we

proceeded to extract maximum (MAX) and

minimum (min) value of the acceleration, their

difference (Diff) and the first derivative of

acceleration (in physics indicated as “jerk”) as an

indication of the smoothness of the performed

movement. The processed signal and the parameters

chosen for the diagnosis are reported in figure 1.

Figure 1: Typical trend of acceleration during the three

repetitions of pivot-shift test. “MAX”, “min” and “Diff”

are the parameters, besides jerk, evaluated in this study.

The automatic data processing is based on the

calculation of Pearson’s correlation coefficient by

means of overlapping a specific template and the

signal recorded. The template was defined as the

average trend manually identifying the repetitions

obtained by the first 35 patients, that were thus used

as reference sample. As known, Pearson’s

correlation coefficient (R, with -1 ≤ R ≤ 1) generally

indicates the strength and direction of a linear

relationship between two variables; in our case the

first variable is the template and the second is the

corresponding window on the processed signal, so

that R can be helpful to recognize the presence of a

trend similar to what represented in the template. For

instance, supposing to have two variables X and Y,

with means

and

respectively and standard

deviation Sx and Sy, n is the number of samples of X

and Y, the correlation R is computed as shown in the

equation 1:

⎟

⎠

⎞

⎜

⎝

⎛

−

⎟

⎠

⎞

⎜

⎝

⎛

−

∑

YYi

XXi

i 1

Equation 1: Calculation of Pearson’s correlation

coefficient (R).

During each iteration of the algorithm, the

evaluation took into account a shear of signal with

the same length of the template sample-by-sample

until the last one.

A threshold for the value of coefficient R over that

the algorithm recognize the PS repetition was

manually fixed to 0.7, on the basis of the reference

sample. Moreover in order to increase the reliability

of the algorithm specific thresholds were defined

relative to the clinical parameters chosen for the

AUTOMATIC SCREENING OF ACCELERATION SIGNAL DURING PIVOT-SHIFT TEST BASED ON PEARSON'S

CORRELATION COEFFICIENT

237

diagnosis, thus avoiding clinically meaningless false

identifications. In particular “MAX” must be ≤ 14.3

m/s^2, “min” must be ≥ 6.5 m/s^2 and “Diff” must

be ≥ 2.2 m/s^2. In figure 2 is reported the template

used for the automatic data processing while flowing

along the signal automatically recognizes the PS

repetitions.

Figure 2: The automatic analysis. In blue, inside the frame,

is represented the window chosen as template which is

flowing along the signal, in green the first end the second

repetition of PS recognized thanks to an high Pearson’s

correlation coefficient associated with the template (R ≥

0.7), while in red is underlined the shear of signal already

processed and in black the signal which is going under

processing.

The automatic method identification performance

was evaluated in terms of percentage sensitivity,

specificity and accuracy, applying the algorithm on

the entire set.

3 RESULTS

Using Pearson’s correlation coefficient, together

with the other mentioned thresholds, provided us a

classification accuracy around 96.7%. The

classification performance measures are presented in

table 1.

Table 1: Classification performance on the test for the

method used for the automatic recognition of the PS test.

% VALUE

Sensitivity

81.9

Specificity

99.3

Accuracy

96.7

For the evaluation of this classification a direct

method was used, i.e. the training set was classified

with a classifier built using the training set data

itself.

4 CONCLUSIONS

The goal of our study was to develop a methodology

able to provide a quantification of the PS test. We

defined therefore a method to easily analyse the

acceleration signal during PS test with the goal to

achieve non-invasive, portable and user-friendly tool

able to provide a quantitative and complete diagnose

of ACL deficient knee. This method used a

predefined template and the computation of

Pearson’s correlation coefficient between this

prototypical pattern and the signal under processing.

In literature there are different methods dedicated to

pattern recognition, even if they are more specific

and require a major computational load.

Being acceleration PS event quite readily

recognizable and identifiable from the remaining

parts of signal, we privileged the easiness of

computation, in order to apply the automatic

recognition also to real-time applications.

A limitation of this study was due to the template

form; its simple and linear waveform is quite easy to

generate also by a mistakenly moved tool. This

conditions could generate false positive (FP) cases.

On the other hand a sudden motion during the test

could falsify the typical trend of the signal,

provoking false negative (FN) cases. Moreover the

coupled loads applied by the surgeon were not

quantitatively controlled and the considered

parameters could be strongly affected.

We are convinced that provide to the surgeons an

accurate description of the movement to perform in

order to realize the pivot-shift test always in the

same way and limit the movements of the tested

limb, usually provoked by the operator to relax the

limb, could improve the performance of the model

regarding false-positive and false-negative.

Further analyses imply to verify the skill, in terms of

sensitivity, specificity and accuracy, of the clinical

method purposed to discern between pathologic and

healthy knee.

ACKNOWLEDGEMENTS

We would like to acknowledge the help of the

anaesthetists and all the staff of the IX Orthopaedic

Division of Istituto Ortopedico Rizzoli, for their

fundamental availability during this study.

BIOSIGNALS 2010 - International Conference on Bio-inspired Systems and Signal Processing

238

In particular we are thankful to Dr. Giovanni

Giordano for his professional support during the

execution of the PS-tests and the whole study.

REFERENCES

Kocher, M.S., Steadman, J.R., Briggs, K., Sterett, W.I. &

Hawkins, R.J., 2009. Relathionships between

objective assessment of ligament stability and

subjective assessment of symptoms and function after

anterior cruciate ligament reconstruction. Arthroscopy

and Sports Medicine, 129, pp.353-358.

Jakob, R.P., Staubli, H.U. & Deland, J.T., 1987. Grading

the pivot shift. Objective tests with implications for

treatment. The Journal of Bone and Joint Surgery,

69(B), pp.294-299.

Lopomo, N., Bignozzi, S., Zaffagnini, S., Giordano, G.,

Irrgang, J.J., Fu, F. & Marcacci, M., 2008.

Quantitative correlation between IKDC score, static

laxity and pivot-shift test: a kinematic analysis of knee

stability in anatomic double-bundle ACL

reconstruction. Operative Techniques in Orthopaedics,

18, pp.185-189.

Kuroda, R., Hoshino, Y., Nagamune, K. & Kubo, S.,

2008. Intraoperative measurement of pivot shift by

electromagnetic sensors. Operative Techniques in

Orthopaedics, 18, pp.190-195.

Musahl, V., Bell, K.M., Tsai, A.G., Costic, R., Allaire, R.,

Zantop, T., Irrgang, J.J. & Fu, F., 2007. Development

of a simple device for measurement of rotational knee

laxity. Knee Surgery, Sport Traumatology, Arthoscpy,

15, pp.1009-1012.

Amis, A.A., Cuomo, P., Rama, R.B.S., Giron, F., Bull,

A.M.G., Thomas, R. & Aglietti, P., 2008.

Measurement of Knee laxity and Pivot-Shift

Kinematics with Magnetic Sensors. Operative

Techniques in Orthopaedics, 18, pp.196-203.

Bleday, R.M., Fanelli, G.C., Giannotti, B.F., Edson, C.J.

& Barrett, T.A., 1998. Instrumented measurement of

the posterolateral corner. The Journal of Arthoscopic

and Related Surgery, 14 (5), pp.489-494.

Diermann, N., Schumacher, T., Schanz, M., Raschke,

M.J., Petersen, W. & Zantop, T., 2009. Rotational

instability of the knee: internal tibial rotation under a

simulated pivot shift test’. Arthroscopy and Sports

Medicine, 129, pp.353-358.

Csintalan, R.P., Ehsa, A., McGarry, M.H., Fithian, D.F. &

Lee TQ., 2006. Biomechanical and anatomical effects

of an external rotational torque applied to the knee.

The American Journal of Sports Medicine, 34,

pp.1623-1629.

Amis, A., Bull, A.M.J., Lie, D.T., 2005. Biomechanics of

rotational instability and anatomic anterior cruciate

ligament reconstruction. Operative Techniques in

Orthopaedics, 15, pp.29-35.

Labbe, D.R., De Guise, J.A., Godbout, V., Fernandes, J. &

Hagemeister, N. Development of an objective

measurement tool for the pivot shift phenomenon of

the knee. 16

Congress of the European Society of

Biomechanics, Lucerne, Switzerland, 2008.

Corthout, J., Van Huffel, S., Mendez, M.O.,Bianchi, A.M.,

Penzel, T. & Cerutti, S. Automatic screening of

obstructive sleep apnoea from the ECG based on

empirical mode decomposition and wavelet analysis.

Annual International IEEE EMBS Conference,

Vancouver, British Columbia, Canada, 20-24 August

2008.

Khandoker, A.H. Karmakar, C.K. & Palaniswami, M.,

2009. Automated recognition of patients with

obstructive sleep apnoea using wavelet-based features

of electrocardiogram recordings. Computer in Biology

and Medicine, 39, pp.88-96.

Colombet, P., Robinson, J., Christel, P., Franceschini, J.P

& Djian, P., 2007. Using navigation to measure

rotation kinematics during ACL reconstruction.

Clinical Orthopaedics and Related Research, 454,

pp.59-65.

Hoshino, Y., Kuroda, R., Nagamune, K., Yagi, M.,

Mizuno, K., Yamaguchi, M., Muratsu, H., Yoshiya, S.,

Kurosaka, M., 2007. In vivo measurement of the

pivot-shift test in the anterior cruciate ligament-

deficient knee using an electromagnetic device. The

American Journal of Sports Medicine, 35, pp.1098-

1104.

AUTOMATIC SCREENING OF ACCELERATION SIGNAL DURING PIVOT-SHIFT TEST BASED ON PEARSON'S

CORRELATION COEFFICIENT

239