Computer-assisted analysis of lower limb geometry: higher intraobserver reliability compared to conventional method

Abstract

Exact radiographic evaluation of lower limb alignment, joint orientation and leg length is crucial for preoperative planning and successful treatment of deformities, fractures and osteoarthritis. Improvement of the accuracy of radiographic measurements is highly desirable. To determine the intraobserver reliability of conventional analysis of lower extremity geometry, 59 long leg radiographs were randomly analyzed 5 times by a single surgeon. The measurements revealed a standard deviation between 0.36° and 1.17° for the angles mLPFA, mLDFA, MPTA, LDTA, JLCA and AMA (nomenclature according to Paley), and 0.94 mm and 0.90 mm for the MAD and leg length, respectively. Computer-assisted analysis with a special software significantly reduced the standard deviation of the mLDFA, MPTA, LDTA, JLCA (each p < 0.001), AMA (p = 0.032) and MAD (p = 0.023) by 0.05–0.36° and 0.14 mm, respectively. Measuring time was reduced by 44% to 6:34 ± 0:45 min (p < 0.001). Digital calibration by the software revealed an average magnification of conventional long leg radiographs of 4.6 ± 1.8% (range: 2.7–11.9%). Computer-assisted analysis increases the intraobserver reliability and reduces the time needed for the analysis. Another major benefit is the ease of storage and transfer of digitized images. Due to the varying magnification factors on long leg radiographs, the use of magnification markers for calibration is recommended.

• Analysis
• alignment
• joint orientation
• intraobserver reliability
• software
• magnification

Introduction

Exact radiographic evaluation of alignment, joint orientation and leg length is crucial for accurate preoperative planning and successful treatment of osteoarthritis [1–3], deformities [4], [5] and fractures [6], [7]. The conventional methods of preoperative planning use tracing paper, transparent films, or original radiographs for the drawings, and a goniometer and long ruler for manual measurements [8–11]. The accuracy of measurement of lower leg geometry depends on several factors such as rotation of the limb [12–14], knee position [9], [15], quality of radiographs [9], and the use of long or short films [3], [16]. One important factor is intraobserver reliability, which has not been analyzed in detail until now [17], [18]. A new software features digital analysis of alignment, joint orientation and leg length on long standing radiographs.

The first aim of this study was to assess the intraobserver reliability using conventional measurement (CM) of the alignment, joint orientation and leg length on long standing radiographs. The second aim was to compare these results to the intraobserver reliability of computer-assisted analysis (CAA) with a special software (mediCAD, Hectec GmbH, Altfraunhofen, Germany). In addition, we compared the time needed for measurement. The third aim was to determine the magnification factor of conventional standardized long standing radiographs, since unknown magnification on the radiographs can cause selection of incorrect implant sizes (which may result in iatrogenic fractures following implantation of an oversized prosthesis [19]), implant misplacement [6], inaccurate correction of deformities, or wrong calculation of the leg length [18].

Materials and methods

Fifty-nine long standing anteroposterior radiographs (130 cm × 36 cm) of 30 patients were analyzed in this study. All patients (13 men, 17 women, average age 46.8 years, range 19–89 years) had been treated between 1998 and 2000 at our institute with unilateral (n = 28) or bilateral (n = 2) locked screw plates (LISS = Less Invasive Stabilizing System, Synthes, Bettlach, Switzerland) because of distal femoral fractures. One radiograph was excluded from the study due to a revision knee prosthesis of the uninjured leg. A total of 59 long standing radiographs (29 bilateral and one unilateral) remained for analysis of the angles mLPFA, mLDFA, MPTA, LDTA, JLCA and AMA according to the system of Paley [9] (Table I, Figure 1). For analysis of the leg length and MAD, 32 radiographs containing LISS plates of known length as a reference for magnification were used.

Figure 1. Joint orientation angles and axis in the frontal plane according to Paley et al. [5]. [Color version available online.]

Table I.  Nomenclature of joint orientation and axis in the frontal plane according to Paley [9].

mLPFA	mechanical Lateral Proximal Femoral Angle: Angle between the mechanical femoral axis and a line from the tip of the greater trochanter to the hip center.
mLDFA	mechanical Lateral Distal Femoral Angle: Angle between the mechanical femoral axis and a tangent through the two most convex distal points of the femoral condyles.
MPTA	Medial Proximal Tibial Angle: Angle between the mechanical axis of the tibia and a line along the flat portion of the subchondral bone of the tibial plateau.
LDTA	Lateral Distal Tibial Angle: Angle between the mechanical tibial axis and a line through the tip of the medial and lateral talus shoulder.
AMA	Anatomical-Mechanical Angle: Angle between the anatomical and mechanical femoral axes.
JLCA	Joint Line Convergence Angle: Angle between the tangent through the two most convex distal points of the femoral condyles and a line along the flat portion of the subchondral bone of the tibial plateau.
MAD	Mechanical Axis Deviation: Distance of the mechanical axis of the leg from the knee center.

All radiographs were taken using a standard protocol [9]. The X-ray tube was positioned 300 cm from the film. The hip and knee joints were fully extended while the patients were full weight bearing on both legs. The X-ray beam was centered at the level of the knee joint with the patella facing directly forward, centered between the femoral condyles. All radiographs showed the whole leg, including the entire hip, knee and ankle joint.

Conventional measurement (CM)

The radiographic films were placed on a typical viewing box with a fluorescent light source, and tracing paper was rigidly fixed over the films to prevent movement during tracing. The contours of the whole femur, tibia and talar dome were exactly drawn with a sharp pencil. The drawings were not made directly on the radiographs, in order to withhold information for the following measurements.

The center of the femoral head was identified with the help of the concentric circles of a goniometer. The knee center was determined by the midpoint of the tibia plateau line, which was drawn from the medial to the lateral border of the tibia plateau. The ankle center was positioned at the midpoint between the edges of the medial and lateral shoulder of the talus [9], [20]. The mechanical axis of the femur and tibia was defined by a line from the center of the femoral head to the knee center and by a line from the knee center to the center of the ankle [9], [21]. The angles were measured manually with a transparent goniometer (1° scale) and a transparent ruler. The leg length was measured from the femoral head center to the ankle center with a transparent ruler. The analysis time was measured from the beginning of tracing to the documentation.

Computer-assisted analysis (CAA)

For computer-assisted analysis, the mediCAD planning software was used (Hectec GmbH, Altfraunhofen, Germany; approved by the FDA). This software features digital analysis of alignment, joint orientation and leg length. The software also features simulation of the correction of deformities, implantation of arthroplasties, and fracture reduction and stabilization. For determination of the magnification factor of the conventional radiographs, the length of the LISS implant served as a calibration reference. The length of the LISS implant was 156 mm (5-hole LISS), 236 mm (9-hole LISS) or 316 mm (13-hole LISS). The contours of the femur, tibia and talus were drawn digitally (Figure 2a). The mLPFA, mLDFA, MPTA, LDTA, AMA, JLCA, MAD and leg length were calculated by the software (Figure 2b).

Figure 2. Digital analysis of leg geometry with the planning software. a) The contours of the femur, tibia and talus are drawn digitally. b) Measurement of the joint orientation angles, alignment and leg length by the software. [Color version available online]

All drawings and measurements of the 59 radiographs were repeated 5 times for the CM and CAA groups. All measurements were performed by a single surgeon to exclude interobserver variability. To eliminate memory effects, the measurements were performed in a random manner on different days. A total of 590 measurements of each parameter (mLPFA, mLDFA, MPTA, LDTA, AMA, JLCA, MAD, and leg length) were performed.

Statistical analysis

The data was implemented in the SPSS software (SPSS 11.5, SPSS Inc., Chicago, IL). The average standard deviation (SD) of the 5 measurements of each parameter was calculated and compared between the CM and CAA group using a paired t-test. A p-level of 0.05 was considered to be statistically significant, and a p-level of 0.001 to be highly significant.

Results

After conventional analysis, the intraobserver standard deviation of the joint orientation angles mLPFA, mLDFA, MPTA and LDTA was 0.65–1.17°. Among these angles, mLPFA showed the highest variability. The SD of the angles AMA and JLCA was 0.36° and 0.77°, and the SD of the MAD and leg length was 0.94 mm and 0.90 mm, respectively (Table II).

Table II.  Standard deviation (SD) after conventional measurement (CM) and computer-assisted analysis (CAA).

	SD CM	SD CAA	SD difference	p-value
mLPFA (°)	1.17	1.12	0.05	0.294
MLDFA (°)	0.65	0.43	0.22	< 0.001
MPTA (°)	0.67	0.47	0.20	< 0.001
LDTA (°)	0.89	0.53	0.36	< 0.001
AMA (°)	0.36	0.29	0.07	0.032
JLCA (°)	0.77	0.58	0.19	< 0.001
MAD (mm)	0.94	0.80	0.14	0.023
Leg length (mm)	0.90	0.96	0.06	0.271

After analysis with the software, the mean intraobserver standard deviation of the joint orientation angles ranged from 0.43° to 1.12°. The intraobserver variability for the angles AMA and JLCA was 0.29° and 0.58°, and that for the MAD and leg length was 0.80 mm and 0.96 mm, respectively.

Compared to conventional measurements, the standard deviation of the AMA and MAD was reduced significantly after computer-assisted analysis (p = 0.032 and p = 0.023). The decrease in SD for the mLDFA, MPTA, LDTA and JLCA was highly significant (p < 0.001). The LPFA and leg length did not reveal significant differences (Table II). In both groups, the smallest variability was measured for the AMA, and the highest variability for the mLPFA. The average time for conventional analysis was 11:46 ± 1:20 min. Using the software, the time for the analysis was reduced significantly to 6:34 ± 0:45 min (p < 0.001) (Table III).

Table III.  Time period for complete analysis of the alignment, axis, joint orientation and leg length.

	CM	CAA	Difference	p-value
Time (min:sec) ± SD (range)	11:46 ± 1:20 (8:17–15:46)	6:34 ± 0:45 (5:00–9:38)	5:11	< 0.001

The mean leg length measured on conventional radiographs was 81.29 ± 5.96 cm. After calibration by the software, a mean leg length of 77.38 ± 4.94 cm was determined (p < 0.001). The average magnification factor of conventional long standing radiographs was 4.6 ± 1.8% (range: 2.7–11.9%) (Table IV).

Table IV.  Leg length before and after calibration with the magnification marker calculated by CAA

CM	CAA	Difference	Magnification
81.29 ± 5.96 cm	77.38 ± 4.94 cm	3.91 ± 2.03 cm	4.6 ± 1.8% (range: 2.7–11.9%)

Discussion

Several factors influence the accuracy of the radiographic analysis of leg geometry. These include the rotation of the limb [12–14], knee position [9], [15], morphological abnormalities [22], the quality of radiographs [9], the use of long or short films [3], [16], and the experience of the investigator [1].

One important factor is the intraobserver reliability. In a study by Wright et al. [12], the intraobserver standard deviation of the anatomical tibiofemoral angle was 0.9°, whereas Ilahi et al. [17] found a mean intraobserver SD of 1.5°. Using different observers and a special software, Prakash et al. [23] determined a maximum mean intraobserver variability for the anatomical tibiofemoral angle of 1.2°.

Whereas most previous publications concentrated only on the anatomical tibiofemoral angle, many authors emphasize the meaning of a more detailed analysis of leg geometry [5], [8], [21]. Sanfridsson et al. [18] analyzed several joint orientation angles and found an intraobserver SD between 0.34° and 2.08°.

In our study, conventional measurement of the angles and distances of Paley's alignment test [9] showed an intraobserver SD between 0.36° and 1.17°. Compared to the results of other authors, the average variability was lower in our study. In comparison to the conventional measurements, the variability of the mLDFA, MPTA, LDTA, AMA, JLCA and MAD was significant lower in the computer-assisted group, with a difference in SD ranging from 0.05° to 0.36°.

These differences between conventional and computer-assisted analysis are relatively minor. Nevertheless, any improvement in accuracy and reproducibility is to be welcomed and is of some clinical relevance. Computer-assisted analyses might also be useful for scientific investigations to detect small differences.

There are several potential explanations for the superior reproducibility of computer-assisted analysis. The software can zoom parts of the radiograph allowing very exact determination of bone and joint surface lines. Tracing paper or transparent films may slide and thus cause inaccurate drawings and decreased reliability. In contrast to the software, the 1° scale of the goniometer used in our study makes it difficult to measure very accurately.

We found the highest intraobserver reliability in the measurement of the anatomical-mechanical axis (AMA) of the femur. The long distances of the anatomical and mechanical axes of the AMA help to prevent inaccurate drawings and enable more accurate measurement. In contrast, the lowest reliability was revealed when measuring the mechanical lateral proximal femur angle (mLPFA). This might be due to the short distance between the hip center and the tip of the greater trochanter. Another reason might be difficulties in determining the exact center of the femoral head, even with the goniometer. Sanfridsson et al. [18] also found the smallest reliability in their study for the neck-shaft angle.

The accuracy of the measurement of distances depends on the magnification of the radiographs. Inaccurate distances can have a significant impact on corrective osteotomies, arthroplasties or fracture stabilization [6], [19]. Conn et al. determined a mean magnification of 18% on radiographs of the hip (range: 6–31%) [19]. Krettek et al. [6] found a mean magnification for femoral radiographs of 9% (range: 1–26%) and for tibial radiographs of 7% (range: 2–10%). We found an average magnification for long standing radiographs of 4.2% (range: 2.7–11.9%). As the magnification varies in spite of standardized techniques, we recommend using a magnification marker for calibration [9], [19], e.g., a radiodense ball fixed laterally to the knee joint. After calibration with the software, the magnification factor is automatically applied to all measurements or implant sizes.

Further advantages of computer-assisted analysis are the reduced planning time, digital storage, and the option of digital transfer. There is no need for sophisticated workstations, only for a simple personal computer, and a scanner if conventional radiographs must be converted into digital radiographs. Disadvantages of the software are the costs arising from the acquisition, and the training period before it enters daily routine.

Computer-assisted analysis increases the intraobserver reliability of the measurement of joint orientation and axis. The software is able to reduce time required for the analysis, provide digital storage, enable digital transfer, and provide digitally scaled implant templates. We recommend using a calibration marker to quantify the varying magnification on long standing radiographs.

Disclosure

None of the authors have received or will receive benefits for personal or professional use from any commercial party related directly or indirectly to the subject of this article.