Use of Inertial Sensors to Predict Pivot-Shift Grade and Diagnose an ACL Injury During Preoperative Testing

Abstract

Background:

The pivot-shift (PS) examination is used to demonstrate knee instability and detect anterior cruciate ligament (ACL) injury. Prior studies using inertial sensors identified the ACL-deficient knee with reasonable accuracy, but none addressed the more difficult problem of using these sensors to determine whether a subject has an ACL deficiency and to correctly assign a PS grade to a patient’s knee.

Hypothesis:

Inertial sensor data recorded during a PS examination can accurately predict ACL deficiency and the PS score assigned by the examining physician.

Study Design:

Cohort study (diagnosis); Level of evidence, 2.

Methods:

A total of 32 patients with unilateral ACL deficiency and 29 with intact ACLs in both knees had inertial sensor modules strapped to the tibia and femur of each limb for preoperative PS testing under anesthesia. Support vector machine (SVM) methods assessed PS grades on the basis of these data, with the examiner’s clinical grading shift used as ground truth. A fusion of regression and SVM classification techniques diagnosed ACL deficiency.

Results:

The clinically determined PS grades of all 122 knees were as follows: 0 (n = 69), +1 (n = 23), +2 (n = 27), and +3 (n = 3). The SVM classification analysis was 77% accurate in correctly classifying these grades, with 98% of computed PS grades falling within ±1 grade of the clinically determined value. The system fusion algorithm diagnosed ACL deficiency in an individual with an overall accuracy of 97%. This method yielded 6% false negatives and 0% false positives.

Conclusion:

This study used inertial sensor technology with SVM algorithms to accurately determine clinically assigned PS grades in ACL-intact and ACL-deficient knees. By extending the assessment to a separate group of patients without ACL injury, the inertial sensor data demonstrated highly accurate diagnosis of ACL deficiency.