Automatic assessment of the knee alignment angle on full-limb radiographs

Abstract

In this study a fully automatic assessment of the knee alignment angles in full-limb radiographs was developed and compared to manual standard of reference measurements in a prospective manner.

The data consisted of 28 knees which were gathered from total-leg radiographs of 15 patients (12 males and 3 females with a mean age of 29.4 ± 6.9 years) consecutively.

For statistical evaluation, a leave-one-out cross-validation was performed. The pattern recognition and consequently the fully automatic assessment were successful in all patients.

The automatically measured angles highly correlated with the standard of reference (r = 0.989). The mean absolute difference was 0.578° (95% CI: 0.399–0.757°). 82% of the angles differed less than 1° from the standard of reference, 46% differed less than 0.5° and 31% differed less than 0.2°. The automatic method showed a high agreement between repeated measurements (+0.515° to −0.429°).

The automatic assessment of alignment angles in full-limb radiographs were equal to the manual assessment. No measurement related user interaction was necessary to achieve results.

Introduction

Anterior–posterior long-leg radiographs are part of a standardized protocol to evaluate axial alignment of the lower limb [1], [2]. The assessment of lower limb alignment is important in orthopaedic surgery, particularly when planning surgery such as total knee arthroplasty or high tibial wedge osteotomy [3], [4], [5].

The knee alignment is of importance in the planning of corrective knee surgery, and the degree of postoperative angulation is known to be directly associated with outcome [6], [7]. In addition, the degree of mal-alignment correlates with the progression of cartilage loss assessed by magnetic resonance tomography [8] and the joint space narrowing measured on radiography, and is therefore of prognostic value in knee osteoarthritis (OA) [9]. The varus knee alignment induces degenerative changes in the medial knee compartment whereas the valgus knee alignment affects the lateral knee compartment. Considering this, and the fact that lower limb alignment can be thought of as a predictor for osteoarthritis, its importance is highlighted. Moreover, the information concerning the alignment of the lower extremity is also significant in the postoperative follow-up.

Until recently, quantification of the knee alignment imposed difficulties due to insufficient reproducibility because of poorly defined landmarks and imprecise measurement techniques [10]. Limited possibilities of patient positioning during set-up, with common factors such as flexion and rotational alterations, frequently influences the apparent alignment on the lower limb radiograph [11], especially in severe cases pre-operatively and after surgery. In addition, the image may be distorted by inclined X-ray beams [12], [13]. In the digital era, assessment of radiographs which are down-scaled in size to fit on review monitors cause loss of precision.

Aside from the acquisition inaccuracies mentioned above, observer variability is a further important factor possibly reducing the accuracy of manual measurements of limb alignment and mainly depends on experience, skill and attitude of the reader. In the past, intra- and inter-reader measurement reliability has been moderate. Even when using the most advanced radiographic positioning protocols, mechanical axis measurements reliability has not proved to be superior to standard clinical readings [10], [14]. Hence, while minor deviations in the mechanical axis can have a considerable impact on the progression of OA, they are usually not detected in standard knee alignment assessment as they are below the limit of detection [10]. For this reason the precise and accurate quantification of the knee alignment angle is highly relevant in clinical practice, but hampered by observer-variability. The purpose of this study was therefore to prospectively test a fully automatic assessment of knee alignment angles in full-limb radiographs by validating its accuracy and precision.