Tilt and Rotation Correction of Acetabular Version on Pelvic Radiographs

Abstract

Anteroposterior pelvic radiographs are the gold standard of imaging for mechanical hip problems. However, correct interpretation is difficult because the projected morphologic features of the acetabulum and nearly all routinely used hip parameters depend on individual pelvic position, which can vary considerably during acquisition. We developed software that recreates the projected acetabular rim and the measured hip parameters as if obtained in a standardized orientation. The vertical and horizontal distances between two easy identifiable points were used as indicators of tilt and rotation. These points were the middle of the sacrococcygeal joint and the middle of the upper border of the symphyseal gap. Calibration of the indicators was achieved by means of serial pelvic radiographs of 20 cadaver pelves. Validation of tilt indicator in 100 patients and a theoretical error analysis revealed that for accurate tilt prediction an additional one-time lateral radiograph of the pelvis is mandatory. The computer-assisted method allows standardized evaluation of anatomic morphologic differences of femoral coverage (dysplasia, retroversion), making their clinical relevance for development of early osteoarthritis more valuable.

In addition to insufficient acetabular coverage (dysplasia), overcoverage of the femoral head has been recognized as a significant precursor of osteoarthritis affecting predominantly young and active adults.^{11,13,14,35,38,40,41} Acetabular overcoverage can lead to symptomatic femoroacetabular impingement resulting in substantial cartilage damage.^10,27,36,38 In this younger population, subsequent treatment is aimed at preservation of the hip such as trimming the acetabular rim or femoral head or both through surgical dislocation of the hip or reorientation of the acetabulum with a periacetabular osteotomy.^{3,10,11,24,38}

The indications for these procedures depend on clinical findings including positive impingement tests and radiographic assessment, which is based predominantly on anteroposterior (AP) radiographs of the pelvis.^{10,11,13,29,35,38} Acetabular retroversion is a typical example of anterior overcoverage of the femoral head, which is visible on an AP radiograph of the pelvis.^13,35,38 It is characterized by an anterior rim projected cranially and more lateral than the posterior rim, which causes a cross-over sign.^35,37,38 Standard evaluation does not focus attention on variable changes in the appearance of morphologic features of the acetabulum and head coverage because of different pelvic tilt orientations.^{1,2,4,23,37,39} Almost all parameters used to commonly describe morphologic features of the acetabulum including the lateral center-edge angle vary considerably with pelvic malpositioning.^4,39,40,43 An increase in pelvic tilt will lead to an increase in the appearance of anterior head coverage and ultimately to an image of pronounced acetabular retroversion and vice versa.^37,42 Therefore, one can at best estimate head coverage and morphologic features of the acetabulum based on plain radiographs of the pelvic without simultaneous knowledge of pelvic tilt and rotation. Because this assessment will substantially influence diagnosis and treatment, more accurate analysis of morphologic features of the acetabulum is mandatory.^10,37

The purpose of this study was to develop a software program that compensates for pelvic malpositioning during acquisition of digital pelvic radiographs and that allows obtaining accurate measurements of the morphologic features of the acetabulum in a standardized neutral orientation. For additional investigation, three questions were formulated: (1) Can pelvic tilt and rotation be estimated with linear distances on an AP radiograph of the pelvis? (2) How accurate is the prediction of tilt and rotation with these distances? and (3) Is there a potential benefit of adding an additional lateral radiograph?

MATERIALS AND METHODS

The projected anterior and posterior acetabular rim can be distinguished by two lines on an AP radiograph of the pelvis. Other than from largely dysplastic hips, any point of these acetabular lines can be assumed to lie on a sphere. Computer software Hip²Norm^© was developed at the MEM Research Center for Orthopaedic Surgery (University of Bern, Bern, Switzerland) that can calculate the virtual three-dimensional shape of the acetabulum respecting the conical projection of the xray beam. As any pelvic tilt or rotation leads to a reproducible change of space orientation of this virtual acetabular model, the software rotates the pelvis into a predefined neutral orientation and displays the projected outline of the acetabular rim, coverage of the femoral head, and the most commonly used radiographic hip parameters in this standardized orientation.

The computerized evaluation of an AP radiograph of the pelvis takes approximately 2 minutes. Briefly, the centers of both femoral heads and acetabula, the outline of the projected anterior and posterior acetabular rim, and the middle of the sacrococcygeal joint and the upper border of the symphysis must be drawn manually using the computer mouse. The inferior margins of the teardrops are used to set the horizontal reference. After computer calculation, the exact orientation of the pelvis (ie, tilt and rotation) during acquisition of the radiograph is shown. Second, the outline of the acetabular rim is displayed on the screen as it would appear in a neutral pelvic tilt and rotational position. Nine established hip angles and indices for evaluation of morphologic features of the hip are computed in the normalized orientation: total femoral head coverage, lateral center-edge angle, acetabular index, ACM angle, femoral head extrusion index, cross-over sign, retroversion index (the relative length of overlap of the anterior rim compared with the entire length of the lateral acetabular opening), posterior wall sign, and anterior center-edge angle.^{17,26,33,35,37,39,43}

Pelvic inclination and the angle between a horizontal line and a line connecting the sacral promontory and the upper border of the symphysis can be measured on a lateral pelvic radiograph (Fig 1).^9,25 A physiologic or neutral pelvic tilt has been defined by various authors as a pelvic inclination of 60°.^7,28,44 Because lateral radiographs of the pelvis are not obtained routinely, we attempted to estimate pelvic inclination and rotation on a plain AP radiograph by defining two points: the upper border of the symphysis and the sacrococcygeal joint. The vertical distance between these points (Distance a) (Fig 2) will directly correlate with pelvic inclination.³⁷ The horizontal Distance b between the center of the sacrococcygeal joint and the middle of the symphyseal gap was used as an indicator for pelvic rotation around the longitudinal axis (Fig 2).

Fig 1.: The illustration shows how pelvic inclination is defined by the angle between a horizontal line and a line connecting the upper border of the symphysis and the sacral promontory. Neutral pelvic tilt is defined by a pelvic inclination of 60°.^7,28,44

Fig 2.: The image shows how Distance a is the indicator for pelvic tilt and is defined by the vertical distance between the upper border of the symphysis (SY) and the center of the sacrococcygeal joint (SCJ) on an AP radiograph of the pelvis. Distance b is the indicator for pelvic rotation around the longitudinal axis and is defined as the horizontal distance between SY and SCJ.

To detect the reproducibility and reliability in measuring Distances a and b, two independent observers measured both distances twice in blinded manner on 50 consecutive radiographs of the pelvis from patients seen in our outpatient clinic. The time between the two measurements was at least 2 weeks. There were 22 male and 28 female patients. Thirty-nine patients had femoroacetabular impingement and 11 patients had developmental dysplasia of the hip. The mean age of the patients was 32.3 ± 9.6 years (range, 16 – 59 years).

Calibration data were obtained in a series of 20 cadaver pelves (10 of each gender) and with a refined experimental set-up in comparison to a pilot study.³⁷ A modified specifically constructed holding device with two radiotranslucent brackets clamped the pelves between the acetabula and allowed tilting around the interacetabular axis and rotation around the longitudinal axis in 1°-steps. Serial digital AP radiographs of the pelvis were taken with the radiograph-tube distance set to be 120 cm, and the central beam was directed to the midpoint between the upper border of the symphysis and a horizontal line connecting both anterior superior iliac spines.³⁷ The pelves were brought into a starting position, defined previously,³⁷ using a gender-dependent mean value of Distance a (males, 32.3 mm; females, 47.3 mm). Each pelvis was tilted forward and backward in 3°-steps with corresponding measurements of this vertical distance by one observer blinded to the radiographs. The central beam position was corrected before acquisition of each radiograph. In the starting position an additional lateral radiograph was taken to measure pelvic inclination. The maximal tilt amplitude of ± 15° was chosen according to the range in variation of pelvic tilt found in the pilot study.³⁷ Analogous analysis of rotation was achieved by 3°-stepwise rotating and documentation of Distance b within a maximal range of ± 9°.

One hundred consecutive conventional AP radiographs with corresponding lateral radiographs were analyzed prospectively to validate the calculation of the absolute value of pelvic tilt with Distance a. Radiographs were taken in 77 patients with femoroacetabular impingement and in 23 patients with developmental dysplasia of the hip with the same technique mentioned above. There were 40 male and 60 female patients with a mean age of 32.1 ± 9.1 years (range, 15.7–59.1 years). One lateral radiograph of each patient was used to correlate the length of Distance a on the AP radiograph with the pelvic inclination as measured on the lateral radiograph. Distance a as an indicator for pelvic tilt was correlated with the real patient’s pelvic inclination on the lateral radiograph of the pelvis.

To estimate the prediction of tilt and rotation by means of Distances a and b under clinical routine conditions, an additional theoretical analysis was done using a previously described formula.^5,6 The influence of the following errors on acquisition and evaluation of the AP radiographs of the pelvis on Distances a and b was investigated: The variability of the pelvic position in the xray system, the variability of the radiograph-tube distance, the error in identifying the symphysis and the sacrococcygeal joint on the plain radiograph, and the variability of the pelvic anatomy. The magnitudes of the former errors were found by empiric assessment of the clinical conditions at exposure or by three-dimensional measurements of the 20 cadaver pelves (Table 1). Burckhardt et al⁶ described the mathematical calculations in a previous report. The outcome parameters, overall standard deviations of a and b, and the partial derivatives of these distances with respect to the error variables, which are measures for the relevance of each source of error, were calculated.

Table 1: Main Results of the Theoretical Error Analysis

The intraobserver and interobserver variability in measuring a and b were assessed using intraclass correlation coefficients (ICC) which were interpreted as follows: ICC < 0.20 = slight agreement; 0.21–0.40 = fair agreement; 0.41–0.60 = moderate agreement; 0.61–0.80 = substantial agreement; and > 0.80 = almost perfect agreement.³² Gender-dependent differences of pelvic inclination in the starting position were analyzed with the Mann-Whitney U test. The correlation of Distance a with pelvic tilt and Distance b with pelvic rotation in the cadaver set-up and the patients’ measurements were evaluated using the simple linear regression model. Significance was defined as p less than 0.05.

RESULTS

The interobserver and intraobserver variances were almost perfect for Distance a and Distance b in all comparisons (Table 2).

Table 2: Intraclass Correlation Coefficients (ICC)

In the 20 cadaver pelves we found a high linear correlation between the change of pelvic tilt and the mean values of Distance a (males, R² = 0.997, p < 0.0001; females, R² = 0.998, p < 0.0001) (Fig 3) and between rotation and the mean values of Distance b (males, R² = 0.996, p < 0.0001; females, R² = 0.998, p < 0.0001) (Fig 4). The mean pelvic inclination measured on a lateral radiograph of the 20 pelves orientated with a Distance a according to the mean value of a normal population was 62° ± 5.6° (range, 55°–73°) for males and 62.2° ± 4.2° (range, 55°–71°) for females.³⁷ We observed no differences between genders (p = 0.46).

Fig 3.: This diagram shows the linear relationship between change of pelvic tilt and the change of Distance a measured on an AP radiograph of the pelvis. Radiographic neutral orientation was chosen according to the mean values for Distance a found in a pilot study (47.3 mm for females, 32.3 mm for males).³⁷

Fig 4.: A–B. These diagrams show the linear relationship between pelvic rotation around the longitudinal axis and corresponding measurements of Distance b for (A) males and (B) females on an AP radiograph of the pelvis.

A moderate relationship was found for the linear correlation between the actual pelvic inclination as measured on the lateral radiograph of the pelvis and Distance a in males (R² = 0.473, p = 0.001) and females (R² = 0.43, p = 0.008) (Fig 5). Therefore, in contrast to the strong correlation between the relative change of Distance a and the change of pelvic tilt, the absolute value of the inclination of an individual pelvis can be moderately estimated.

Fig 5.: A–B. The graphs show the correlation between the real pelvic inclination angle and the vertical distance between the middle of the upper border of the symphyseal gap (SY) and the middle of the sacrococcygeal joint (SCJ). Distance a is given for (A) males and (B) females.

The standard deviations for pelvic tilt were 9.2 mm for males and 7.5 mm for females (ie, Distance a varies in female pelves with a probability of 68% of as much as ± 7.5 mm from the expected value). For pelvic rotation, the respective values are 2.1 mm for males and 2.4 mm for females. For determination of the error bounds of tilt and rotation angle, it was assumed that the probability of the real angle lying within the bounds has to be 95%. This was found to be the case at bounds of ± 8.5° and ± 6.4° for tilt and ± 1.9° and ± 2.1° for rotation in males and females. That means that the estimation of pelvic rotation based on Distance b can be assumed to be relatively accurate, whereas the tilt only can be estimated with a significant error when using only Distance a as the indicator.

The percentage of each potential error and the overall standard deviation of Distances a and b depend on the product between the partial derivative of Distance a or b with respect to the error variable and the standard deviation of the latter (Table 1, Column 2). The individual morphologic features of the sacrum are the most important factors impeding an accurate prediction of tilt. The variability of the pelvic x-y position in the xray system, ie, the positioning of the central beam, also plays a significant role.

Two clinical examples are presented from the patient database to show the potential of the software. In a first case, the pelvis had an inclination of 50°, which means a tilt of −10° and no rotation to either side (Fig 6). In this position, acetabula had a positive cross-over sign as an apparent acetabular retroversion. After computer-calculated display of the pelvis in a neutral position with a pelvic inclination of 60° and neutral rotation, the cross-over signs on both sides disappeared, showing normal anteversion of the acetabula.

Fig 6.: The image shows apparent retroversion. Based on the AP and lateral radiographs of the pelvis, the computer analysis revealed that orientation of the pelvis during acquisition of the radiograph was in 50° pelvic inclination with no rotation to either side. Positive cross-over signs on both sides indicate acetabular retroversion. The computerized standardization to the neutral pelvic position (small windows) revealed disappearance of the cross-over signs and normal acetabular anteversion.

In a second case, pelvic inclination measured on a lateral radiograph was 58°, representing a pelvic tilt of −2° (Fig 7). The patient had a previous periacetabular osteotomy on the left side for developmental dysplasia of the hip. Cranial acetabular retroversion was visible on the left side which disappeared after computerized standardization to neutral orientation. Instead, the calculated computer simulation revealed an abnormal oblique roof orientation indicating acetabular dysplasia and cranial retroversion on the right side, which was not evident on the initial radiograph.

Fig 7.: This patient had a periacetabular osteotomy on the left side for developmental dysplasia of the hip. The estimated rotation angle was 22° to the left and the measured tilt angle was −2° (pelvic inclination, 58°). On the nonstandardized radiograph, a cranial acetabular retroversion is visible which disappears after computerized standardization to the neutral orientation (small windows). Instead, the simulated acetabular configuration on the right side revealed a dysplasia with cranial retroversion that is not evident on the original radiograph.

DISCUSSION

In addition to modern radiographic imaging techniques such as magnetic resonance imaging or computed tomography (CT), plain radiography remains the essential, easy, and cost-effective technique for initial evaluation and interpretation of abnormalities of the hip.^12,13,30,31 The radiographic finding of a retroversion sign of the acetabulum with the anterior border of the projected rim crossing the posterior border is an indicator for a prominent anterosuperior acetabular rim predisposing to femoroacetabular impingement.^{11,13,35,38,38} A significant association between retroversion and hip osteoarthritis was described in a study of nonstandardized AP radiographs of the pelves of 131 patients.¹³ However, considering that pelvic tilt and rotation were not defined on these radiographs, the accuracy of the evaluation of the shape of the pelvic rim from a morphologic standpoint is questionable. Other than intraindividual variations of pelvic positioning on a radiograph table, a positive cross-over sign can be caused by additional factors such as an altered posture or change in lumbar lordosis attributable to inflammatory or degenerative diseases, and is not just a consequence of the morphologic features of the hip. Our suggested method provides a tool to distinguish between anatomic and functional causes. In addition, it can be used to describe the influence of pelvic orientation on the appearance of morphologic features of the hip. Therefore, it provides a more strict anatomic-based analysis of morphologic features of the hip and allows quantifying femoral head undercoverage and overcoverage. The prediction of pelvic inclination with Distance a was not adequately accurate: an initial lateral radiograph was necessary for calibration. Because the linear relationship between Distance a and tilt changes, any other radiograph of the same patient can be adjusted to a neutral position in 60° inclination even when pelvic tilt has changed. Pelvic rotation can be compensated with using Distance b.

Some restrictions apply to this method. A prerequisite for a reliable analysis of morphologic features of the hip with the Hip²Norm^© software is acquisition of the AP radiograph of the pelvis in a standardized manner. The software analysis also requires a spherical acetabulum and clearly recognizable, projected acetabular rims. The software is not applicable to dysplastic hips with an obviously oval-shaped configuration.

Various methods for determination of femoral head coverage have been described (Table 3). A special conventional radiographic technique with a modified inlet view was proposed, again with the disadvantage of disregarding individual pelvic tilt.¹⁹ Alternatively, CT-based measurements could provide greater anatomic details, but they do not solve the problem of different individual pelvic orientations during data acquisition.^18,21,34 In addition, routine CT-scans for detecting patients with femoroacetabular impingement would be time-consuming and lead to an increased diagnostic workload.

Table 3: Methods of Determination of Acetabular Coverage

Our concept of calculating coverage of the femoral head is based on the geometric model of a spherical joint and the use of an AP radiograph of the pelvis. The basic principle has been used for analyses of head coverage with templates or with other computer software.^{8,15,16,22,23} Each method has drawbacks, such as ignoring individual pelvic tilt and conical projection of the xray beams.^8,15,16 A parallel projection implies a discrepancy in the appearance of the acetabular rim compared with a standard radiograph of the pelvis, on which the anterior rim is seen more prominently because it lies closer to the source of the xray beam.³⁰ In some studies, a parameter was introduced to reduce calculation errors attributable to varying pelvic tilt orientations.^19,42,43 The parameters included the ratio of the height of the obturator foramen to the interteardrop distance or the distance between the symphysis and a line connecting the bilateral centers of the femoral head.^20,22,23 However, in contrast to our study, none of these parameters has been validated in larger patient series or by theoretical error analysis. Therefore, accuracy of the correlation of the chosen parameters with actual pelvic tilt is unknown. In addition, no parameter was introduced to compensate for measurement errors because of pelvic rotation around the longitudinal axis.

Based on a cone projection model, the Hip²Norm^© software offers a method to compensate for tilt and rotation by means of the two easy identifiable distances (a and b) as indicators for tilt and rotation. Both parameters were reliable indicators for relative changes in pelvic tilt or rotational positions. However, only Distance b is a reliable indicator for determination of the absolute value of pelvic rotation, whereas determination of Distance a on an AP radiograph of the pelvis is not accurate enough for determination of the absolute pelvic tilt position. The assumed reason for this inaccuracy is the variation of individual morphologic features of the sacrum. Therefore, the proposed solution is one-time use of an additional lateral radiograph to calibrate Distance a as a parameter for pelvic tilt in each individual. Once the individual correlation has been established, the tilt of any subsequent or previous radiograph of the pelvis of the same patient can be calculated precisely in prospective and retrospective studies because of the intraindividual linear correlation between Distance a and pelvic tilt.

The advantage of this computer-based method is that morphologic features of the hip can be evaluated on a morphologic basis. This will help to eliminate the influence of individual posture on the commonly used radiographic parameters for the hip. Our method also provides a tool for standardized quantification of acetabular coverage. This will allow a more precise method for definition of normal, undercoverage, and overcoverage of the femoral head. The computer-based evaluation is being used in clinical practice in our department and to determine the difference in the appearance of morphologic features of the hip compared with conventional measurements in a larger patient series with femoroacetabular impingement. The data will be used to allow determination of the clinical significance and for a thorough validation of this method.

Acknowledgment

We thank Dr. Morteza Kalhor for help in measuring intraobserver and interobserver agreement.