DOS 2016

DOS Kongressen 2016 ·

147

A novel clinical method for non-invasive quantification

and grading of pivot-shift test

Emil T. Nielsen, Michael S. Andersen, Ole G.Sørensen, Sepp de Raedt, Maiken

Stilling

Orthopedic Research Department, Aarhus University Hospital; Department of

Mechanical Engineering and Manufacturing, Aalborg University; Department

of Sportstraumatology, Aarhus University Hospital; , Nordisk Røntgen Teknik;

Department of Clinical Medicine, University of Aarhus

Background:

Anterior cruciate ligament (ACL) injury may be complicated with

extrinsic ligament injury such as injury to the anterolateral ligament (ALL), which

may increase rotational instability. The pivot-shift (PS) test dynamically repro-

duces knee rotational instability, and positive tests correlate with patients’ sub-

jective experience of knee stability, reduced sports activity, and risk of early

gonarthritis. However, the PS grading is poorly repeatable between clinicians.

Purpose / Aim of Study:

To develop an objective grading system for the PS

test that screened for human errors.

Materials and Methods:

One examiner graded PS tests performed on eight

cadavers exposed to five successive ligament situations: intact, ACL lesion,

ACL+ALL lesion, ACL reconstruction, and ACL+ALL reconstruction. Tibial kin-

ematics were assessed using an inertial measurement unit (IMU) and dynamic

radiostereometry (dRSA) to evaluate the accuracy of the IMU. An automatic

screening algorithm using IMU-features approved 95 PS tests (training: n=76,

evaluation: n=19). Based on IMU-features, four different artificial neural net-

works (ANNs) were developed and trained to grade individual PS tests using the

clinical grades (0,1,2,3) given by the examiner as a gold standard.

Findings / Results:

The RMSE comparison of the IMU and dRSA showed no

difference (p>0.61) between rejected (n=18) and approved (n=14) PS tests.

The automatic screening algorithm correctly categorized 97% of these 32 PS

tests. The two ANNs that used a combined-average strategy had the best ac-

curacy of 84% for grading the 19 PS tests.

Conclusions:

ANNs have a great potential for objective individual grading of PS

tests, and further it is a low-cost and user-friendly method. Following ongo-

ing in-vivo testing and calibration, it may be used for clinical individual rotation

instability grading in patients with knee injuries.

No conflicts of interest reported

98.