Musculoskeletal Imaging • Original Research Guggenberger et al. Assessment of Lower Limb Length and Alignment Using Different Modalities
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Musculoskeletal Imaging Original Research
Assessment of Lower Limb Length and Alignment by Biplanar Linear Radiography: Comparison With Supine CT and Upright Full-Length Radiography Roman Guggenberger 1 Christian W. A. Pfirrmann1 Peter P. Koch2 Florian M. Buck1 Guggenberger R, Pfirrmann CWA, Koch PP, Buck FM
Keywords: biplanar linear x-ray scanner, limb alignment, limb length, supine CT, upright full-length radiography DOI:10.2214/AJR.13.10782 Received February 17, 2013; accepted after revision April 24, 2013. 1 Department of Diagnostic and Interventional Radiology, University Hospital Zurich, Raemistrasse 100, Zurich 8091, Switzerland. Address correspondence to R. Guggenberger ([email protected]). 2 Department of Orthopedic Surgery, Balgrist University Hospital, Zurich, Switzerland.
WEB This is a web exclusive article. AJR 2014; 202:W161–W167 0361–803X/14/2022–W161 © American Roentgen Ray Society
OBJECTIVE. The purpose of this article is to compare lower limb length and alignment measurements on supine CT, upright full-length radiography, and 3D models based on upright biplanar linear radiography. SUBJECTS AND METHODS. This study involved 51 consecutive patients (22 men and 29 women; mean age, 68.8 years; range, 43–92 years) who were scheduled for total knee replacement. Lower limb length and alignment angle were measured on CT, upright fulllength radiography, and 3D models based on biplanar linear radiography with standard and composed leg methods by two independent readers. Descriptive statistics of each modality were calculated. Measurements of different modalities were compared by paired Student t tests. Agreement between readers and modalities was assessed by Bland-Altman analyses. RESULTS. Mean (± SD) limb lengths were 783 ± 56.1 mm (range, 639–927 mm), 785 ± 53.0 mm (range, 655–924 mm), 780 ± 55.4 mm (range, 633–921 mm), and 783 ± 55.9 mm (range, 636–924 mm) for CT, upright full-length radiography, and 3D models based on biplanar linear radiography standard and composed leg measurements, respectively. Mean alignment angles were 2.3° ± 5.5° (range, −12° to 20°) for CT, 2.5° ± 6.7° (range, −17° to 18°) for upright full-length radiography, and 3.4° ± 6.6° (range, −14° to 18°) for 3D models based on biplanar linear radiography. No significant differences among modalities for mean limb length were found when using composed leg measurements in biplanar linear radiography. Very small but significant mean differences in angle measurements were seen for CT (−1.1° ± 2.5) and upright full-length radiography (−0.9° ± 3.1) compared with biplanar linear radiography. Bland-Altman analyses showed no significant differences between readers, with the highest agreement for biplanar linear radiography length measurements. CONCLUSION. Measurements on 3D models based on upright biplanar linear radiographs allow lower limb length and alignment angle measurements that are interchangeable with supine CT scans and upright full-length radiographs but with superior interreader agreement.
D
ifferences in limb length should be reliably quantified before knee replacement surgery, allowing intraoperative correction and symmetric postoperative length of the lower limbs [1]. In addition, a physiologic alignment should be restored. Postoperative varus or valgus deformity of more than 3° has been shown to markedly reduce the durability of total knee replacements and has been associated with a higher rate of surgical revision procedures [2]. Despite some controversy in the literature, a neutral postoperative limb alignment seems to be beneficial for long-term outcome after total knee prosthesis implantation [3, 4]. Today, several imaging modalities allow length and alignment measurements of the lower limb in either a supine or upright weight-
bearing position. The scout view of a supine CT scan may serve as a full-length radiograph but can be acquired only in a non-weight-bearing position. Conventional upright full-length radiography allows weight-bearing upright imaging of the patient but, analogous to supine CT scans, may be associated with measurement errors in cases of rotational misplacement or malpositioning of the lower extremities at the time of image acquisition [5–7]. To assess reliably length and alignment angle of the lower limb, full extension of the knee joint is required. This, however, may be impossible in cases of severe osteoarthritis or contractures of the extensor muscles [8]. Upright biplanar linear radiography with generation of 3D models based on simultaneously acquired frontal and lateral projections
AJR:202, February 2014 W161
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Guggenberger et al.
Fig. 1—52-year-old woman with severe osteoarthritis of knee. For all modalities (supine CT scans, upright full-length radiographs, and 3D models based on biplanar linear radiography), lower limb length was measured between center of femoral head and center of superior contour of talar trochlea (left leg). Alignment angle was defined by line through center of femoral head and intercondylar notch of femur and line between intercondylar eminence of tibia and center of ankle joint (right leg). Positive alignment values signified varus alignment, and negative values signified valgus alignment.
allows both upright imaging under weightbearing conditions and the depiction of 3D structures with consecutive compensation of extension deficits in the knee joints or rotation or projection differences occurring in mere 2D projection modalities. Measurements of the lower limb from 3D models based on biplanar linear radiography have already been shown to be comparable to 3D measurements based on CT in pediatric patients [9] and have shown increased accuracy for hip-knee-ankle angle measurements compared with 2D in ex vivo and in vivo conditions [10, 11]. Measurements from 3D models of the legs allow measurement of the limb length in 3D with and without taking account of an extension defi-
W162
Fig. 2—52-year-old woman with severe osteoarthritis of knee. A–C, Supine CT scout view (A), conventional upright full-length radiograph (B), and upright biplanar linear radiograph (C) of lower limbs are shown. Note slight pelvic rotation to right (A) and stitching artifacts from postprocessing (slight dark horizontal bands, B) whereas homogeneous depiction quality is perceived in panels A and C. Upright linear radiographs from anteroposterior (C) and lateral (not shown) projections were used for generation of 3D models. Measured limb length and alignment angles of both legs in this patient ranged between 721 and 738 mm and 0°–2° valgus for different modalities.
cit in the knee joint. Thus, postoperative limb length after implantation of total knee prostheses, which restores full extension in the knee joint, can be estimated. Biplanar linear radiography has therefore gained increasing attention in recent years in both the orthopedic and radiologic communities and may potentially replace traditional modalities for the assessment of lower limb geometry (i.e., length and alignment angle). However, the question arises whether limb length and alignment measurements using 3D models from upright biplanar linear radiographs are interchangeable with measurements on supine CT scans and weight-bearing upright full-length radiographs and whether measurement reliability is superior in patients with severe osteoarthritis of the knee joint. At present, to our knowledge, no comparison has been made between 3D models from upright
biplanar linear radiography and projectional 2D modalities in the supine (CT scan) or weight-bearing (upright full-length radiography) position under in vivo conditions. Thus, the purpose of our study was to compare lower limb length and alignment angle measurements using supine CT scout views, weight-bearing upright full-length radiography, and 3D models based on biplanar linear radiography. Subjects and Methods Institutional review board approval was obtained before the start of the study. In total, 59 patients with severe osteoarthritis of the knee scheduled for a total knee replacement were prospectively included after giving their informed consent. Eight patients had to be excluded from data analysis because of missing scaling items or insufficient image quality of upright full-length radiographs. Patients younger than 18
AJR:202, February 2014
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Assessment of Lower Limb Length and Alignment Using Different Modalities years, patients with Legg-Calvé-Perthes or developmental dysplasia of the hip, and patients who were unable to stand in an upright position would have been excluded also. Eventually, 51 patients (22 men and 29 women; mean age, 68.8 years; range, 43–92 years) were included in the final data analysis. As a routine preoperative imaging workup, all patients were scheduled for a CT scan of the hip joints, knee joints, and ankle joints for the production of patient-matched cutting blocks for knee prosthesis implantation surgery. The scout views of these CT scans were used for this study. In addition, an upright full-length radiography and a biplanar linear radiography were performed the same day.
CT Scans CT scans were performed using a 64-MDCT unit (Brilliance 64, Philips Healthcare). The patient was scanned in the supine position with the feet entering the gantry first. Scans included the entire lower limb from the iliac crests to the soles of the feet. Both legs were composed as much as possible, and care was taken to stabilize them in a neutral position without any rotation. Ankle joints were flexed in 90°, and the toes were pointed toward the ceiling. The anteroposterior scout view was acquired using the following settings: scan length, 120 cm; tube voltage, 80 kV; tube current, 300 mA; FOV, 500 mm; and matrix, 128 × 512.
Fig. 3—52-year-old woman with severe osteoarthritis of knee. A and B, Semiautomated generation of 3D model is based on matching of 3D model of femur and tibia with osseous contours of lower limb on frontal and lateral linear radiographs (A). Software requires identification of predefined osseous landmarks (e.g., femoral head, femoral condyles, tibial head, and medial and lateral malleolus) to generate 3D model. This model is then fine-tuned manually. Eventually, 3D limb lengths and limb alignments angle are calculated automatically by software (B).
A
Upright Full-Length Radiography Upright full-length radiography was performed on a fully digital radiography system (Ysio, Siemens Healthcare). Different preset tube voltages were chosen by the radiology technician according to the patient’s stature and weight (small, 75– 85 kV; medium, 77–90 kV; large, 77–96 kV), and tube currents were set according to automatic exposure measurements. A scaling item (a small radiopaque sphere with a diameter of 2.5 cm) was positioned next to the right lower leg before image acquisitions and was used for image calibration.
Biplanar Linear Radiography Anteroposterior and lateral linear radiographs were simultaneously acquired using a biplanar radiography system (EOS, EOS Imaging). The patient was centered in the upright position on the scanner platform. The scan included the whole lower limb from the hip joint to the ankle joint. Tube voltage of the x-ray tubes was chosen by the radiology technician among three different presets, according to the patients’ weight and stature (small, medium, and large; 80–104 kV with 200– 320 mA for anteroposterior projection and 100– 120 kV with 200–320 mA for lateral projection). A translational speed of the tubes of 5 was chosen from the vendor-specific scale, ranging from 1
B AJR:202, February 2014 W163
Guggenberger et al.
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
(fast) to 8 (slow). After scanning, the images were sent to an EOS workstation for secondary reconstruction and generation of 3D models.
Postprocessing and Measurements CT scout view and full-length radiography—The measurements were performed by two independent musculoskeletal radiologists with 3 and 7 years of experience. Lower limb length was defined as the distance from the center of the femoral head to the center of the superior contour of the talar trochlea. Lower limb alignment angle was defined by a line through the center of the femoral head and the intercondylar notch of the femur and a line between the intercondylar eminence of the tibia and the center of the ankle joint [12] (Figs. 1 and 2). Biplanar linear radiography—Biplanar linear radiographs were postprocessed using a vendorspecific workstation (sterEOS, EOS Imaging) by two independent radiology technicians with special training in 3D reconstruction based on biplanar linear radiography. The user of the reconstruction software manually marks predefined osseous landmarks, such as the femoral head, femoral condyles, tibial head, and medial and lateral malleolus. Then the software semiautomatically matches a 3D model of a femur and a tibia to the contour of femur and tibia on the frontal and lateral radiographs. On the basis of this 3D model, measurements of limb length and limb alignment are automatically calculated (Fig. 3). Limb length is calculated as the distance between the center of the femoral head and the center of the superior contour of the talar trochlea in 3D models (standard limb length measurement). In addition, there is another way to measure limb length, with the software taking into account extension deficits, where the limb length is calculated by adding up the 3D lengths of the femur and tibia. We call this measurement method the “composed leg” limb length measurement. Limb alignment angles, however, remain unchanged.
Statistical Analysis All calculations were performed using commercially available software (SPSS Statistics, version 19, IBM). Descriptive statistics (mean [± SD] and range)
were calculated for lower limb length and alignment of the three different modalities and both readers. Interreader agreement and agreements between modalities were assessed by Bland-Altman analysis and paired two-tailed Student t testing. Bland-Altman plots were generated to graphically depict scatter of measurement differences. A p value less than 0.05 was considered statistically significant.
Results Lower Limb Length Mean limb lengths were 783 ± 56.1 mm (range, 639–927 mm) for CT scout views, 785 ± 53.0 mm (range, 655–924 mm) for upright full-length radiographs, and 780 ± 55.4 mm (range, 633–921 mm) for 3D models based on biplanar linear radiographs (Table 1). Composed leg limb length measurements revealed a mean limb length of 783 ± 55.9 mm (range, 636–924 mm). Small statistically significant (p < 0.001) differences were seen among the different modalities for lower limb length measurements. Limb lengths measured using 3D models based on biplanar linear radiographs were significantly shorter than measurements on CT scans and upright full-length radiographs. No differences were seen between limb length measurements on CT scans and upright full-length radiographs (p = 0.057). On average, limb length measurements using 3D models based on biplanar linear radiographs were 2.7 ± 5.5 mm shorter than measurements on CT scans and 5.4 ± 13.1 mm shorter than measurements performed on upright full-length radiographs. Regarding the composed leg measurement method, no statistically significant differences were seen in comparison with CT scans and biplanar linear radiographs (Table 2). Lower Limb Alignment Mean alignment angles were 2.3° ± 5.5° (range, −12° to 20°) for CT scout views, 2.5° ± 6.7° (range, −17° to 18°) for upright full-length radiographs, and 3.4° ± 6.6° (range, −14° to 18°) for 3D models based on biplanar linear radiographs (Table 1). Small statistically signifi-
cant (p < 0.001) differences were seen among the different modalities for angle measurements. Compared with measurements using 3D models based on biplanar linear radiographs, angles measured on CT scans were 1.1° ± 2.5° smaller and measurements on upright fulllength radiographs were 0.9° ± 3.1° smaller. No differences were seen for alignment measurements between CT scans and upright fulllength radiographs (p = 0.504) (Table 2). Intermodality Agreement Bland-Altman analyses for comparisons of different modalities showed a constant range of measurement differences and scatter with increasing average for both length and alignment angle measurements. Consistent variability of measurements across all graphs was seen. Further results concerning specific patient subgroups (varus alignment, valgus alignment, varus angle > 10°, and valgus angle > 10° as measured on upright fulllength radiography) are shown in Table 2 and Figures 4 and 5. Interreader Agreement Interreader agreement as calculated with Bland-Altman analyses was high and showed no significant differences between both readers in all modalities. The smallest mean lower limb length difference and range between both readers were seen in measurements on 3D models based on biplanar linear radiographs using the composed leg measurement (mean difference, −0.5 ± 3.4 mm; range, 8 mm) (Table 3). Discussion Despite small statistically significant differences, we found overall clinically comparable lower limb length and alignment angle measurements on supine CT scans, weight-bearing upright full-length radiographs, and biplanar linear radiographs in this study. Although patient posture seems to not significantly affect both measures, Bland-Altman analyses detected small but significant differences between 3D models from weight-bearing bipla-
TABLE 1: Descriptive Statistics of Mean Limb Length and Alignment Measurements for Different Imaging Modalities and Both Readers Reader 1 Imaging Modality
Reader 2
Limb Length (mm)
Limb Alignment Angle (°)
Limb Length (mm)
Limb Alignment Angle (°)
Supine CT scout view
783 ± 55.8
2.3 ± 5.5
782 ± 56.7
2.4 ± 5.5
Upright full-length radiographs
785 ± 53.6
2.5 ± 6.9
786 ± 52.7
2.6 ± 6.7
Upright biplanar linear radiographs
779 ± 55.3
3.5 ± 6.7
781 ± 55.7
3.4 ± 6.6
Composed limb length measurement
783 ± 55.8
784 ± 56.0
Note—Data are mean ± SD.
W164
AJR:202, February 2014
Note—Bold type indicates a statistically significant difference (p < 0.05).
−3, 1
4 14
−3, 11 −5, 3
8 26
−8, 18 −8, 18
26 5
1, 4 −6, 2
8 7
−2, 5 −7, 8
15 14
−7, 8 2, 7
9 14
−12, 2 −1, 8
8 22
−18, 5 −18, 8
25 Range
Minimum, maximum
−0.9 ± 3.1 −0.6 ± 3.7 −1.6 ± 1.6 −1.7 ± 3.7 −1.1 ± 1.8
−1.5 to −0.3 −1.5 to 0.3 −2.1 to −1.0 −0.6 to 4.1 −3.9 to 1.6 0.9–4.3
2.6 ± 1.1 −2.5 ± 2.2 0.3 ± 1.9 −1.9 ± 2.4
−1.6 to −0.6 −2.5 to −1.3 −0.3 to 1.0 −3.9 to −1.0 0–7.6
−4.2 ± 3.6 −3.8 ± 2.4 −1.1 ± 2.5
−6.4 to −2 95% CI
1.9 ± 1.9 −0.2 ± 3.4 −1.3 ± 3.5
−0.9 to 0.4 −2 to −0.4
Mean difference ± SD
Composed leg length
Lower limb alignment angle (°)
1.2–2.6
27 47 43 83 83 12 18 35 18 35
−9, 18
25 38 32 87 87 15 23 35 39 40 33 60 37 81 81 Range
−5, 20 −19, 19
−30, 17 −15, 28
−11, 22 −52, 36
−55, 28 −55, 28
−52, 36 1, 14
−2, 10 −10, 8
−4, 19 −21, 14
−26, 10 −10, 8
−20, 19 −21, 19
−26, 10
−15, 18 −22, 38 −19, 18 −31, 51 −31, 51 Composed leg length
Minimum, maximum
−7 to 10 −7.4 to 24 1.7–7.3 2.2–9.5 2.8–7.9 −3 to 15 −0.1 to 8.6 0.1–4.5 1.7–4.1 1.6–3.8
−1.5 to 0.2 −3.4 to 1.6 −3.7 to 2.5 −5.1 to 10.6 −0.8 to 4.6 −1.5 to 6.0 −2.0 to 4.6 −13 to 7 −12 to 23 Composed leg length
95% CI
−5.4 to 0.1 −6.7 to 0.9 −5.6 to 1.2 −9.3 to 15 −25 to 20
−1.7 to 0.3
−3.3 ± 16 5.4 ± 11
1.6 ± 13 −8.4 ± 10 4.5 ± 8
1.3 ± 9 2.3 ± 16
5.8 ± 15 5.4 ± 13
1.9 ± 14 2.8 ± 4.9
5.8 ± 5.5 4.4 ± 6.7 2.3 ± 6.3
−0.9 ± 7.3 −0.6 ± 4.9 −0.6 ± 3.7
2.9 ± 5.1 2.7 ± 5.5 −2.7 ± 13.9 −2.9 ± 15.6 −2.2 ± 9.8 −2.8 ± 18.9 −2.6 ± 14
Composed leg length
Mean difference ± SD
Lower limb length (mm)
Limb, Measurement
−0.7 ± 5.1
Valgus > 10° (n = 4) Varus > 10° (n = 12) Valgus (n = 34) Varus (n = 68) All (n = 102) Valgus > 10° (n = 4) Varus > 10° (n = 12) Valgus (n = 34) Varus (n = 68) All (n = 102) Valgus > 10° (n = 4) Varus > 10° (n = 12) Valgus (n = 34) Varus (n = 68) All (n = 102)
Upright Full-Length Radiographs vs Upright Biplanar Linear Radiographs Supine CT vs Upright Biplanar Linear Radiographs Supine CT vs Upright Full-Length Radiographs
TABLE 2: Bland-Altman Analyses of Measurement Differences in Different Modalities for Limb Length and Alignment Measurements in All Patients and Subanalyses for Patients With Varus, Valgus, Varus > 10°, and Valgus > 10°
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Assessment of Lower Limb Length and Alignment Using Different Modalities nar linear radiography and both other modalities. However, these differences were in the order of magnitude of the measurement accuracies themselves. Interreader agreement was high for all modalities, with the highest agreement seen in lower limb length measurements using 3D models based on biplanar linear radiographs. Upright full-length radiography offers fast depiction of the lower limb in an upright patient position. Technically, it consists of three projections at different angulations that are stitched together and visually corrected by the technician to form a single image. Hence, upright full-length radiographs may, in addition to the impact of posture, be associated with a certain distortion effect when compared with supine CT scans, leading to the observed differences in severe misalignment angles [5–7]. Similarly, low-dose scout views in CT scans are also a projection of 3D structures onto a 2D plane but performed with the patient in a non-weight-bearing supine position. In contrast to upright full-length radiography, CT scout views are not associated with distortion artifacts due to possible stitching errors. Therefore, observed mean differences between both modalities for lower limb length and alignment angles were small (average, 2.7 mm and 0.2°, respectively) and may be considered clinically irrelevant [1, 2]. Interestingly, larger differences of limb alignment angle measurements were seen when comparing supine CT scans to weight-bearing upright full-length radiography in patients with varus or valgus alignment of more than 10° (average, −4.2° for varus angles > 10° and −3.8° for valgus angles < −10°). This increase in difference between both modalities may be caused by increased lower limb misalignment angles under upright weight-bearing conditions during upright full-length radiography imaging. Therefore, marked varus or valgus deformity posture during imaging seems to increasingly affect measured lower limb alignment angles. Upright biplanar linear radiography with reconstruction of 3D models based on simultaneously acquired contiguous frontal and lateral projections combines the advantage of a physiologic upright weight-bearing position during imaging with the possibility of 3D model generation at a low radiation dose [9, 13]. To be able to generate exact 3D models, the patient needs to be positioned in a small step position to allow delineation of the posterior contour of the femoral condyles of both legs on the lateral view. Overall, limb length measurements were comparable among different modalities, and mean differences were small from a clinical point of view. Limb length measurements on 3D models using standard measurements were, however, significantly smaller compared with supine CT scans and weight-bearing upright full-length radiography. These mean differences increased in patients with severe varus and valgus deformity (average, 4.4 and 5.8 mm, respectively) comparing biplanar linear radiography to CT. Limb length is measured on 3D models as the distance from the center of the femoral head to the center of the upper ankle joint. Thus, flexion of the knee joint at the time of image acquisition leads to a falsely short limb length measurement in these 3D models. The composed leg mea-
AJR:202, February 2014 W165
0 Bias
–20
Lower Limit (95% CI)
Upper Limit (95% CI) Bias
0
Lower Limit (95% CI)
–20 600 700 800 900 1000 Average Lower Limb Length (mm)
Upper Limit (95% CI)
20
Bias
0 –20
Lower Limit (95% CI)
600 700 800 900 1000 Average Lower Limb Length (mm)
C
20
Upper Limit (95% CI)
0
Bias Lower Limit (95% CI)
–20
Difference (mm)
B
Difference (mm)
A
20
Difference (mm)
Upper Limit (95% CI)
Difference (mm)
Difference (mm)
20
600 700 800 900 1000 Average Lower Limb Length (mm)
600 700 800 900 1000 Average Lower Limb Length (mm)
D
Upper Limit (95% CI)
20 0
Bias
–20
Lower Limit (95% CI)
600 700 800 900 1000 Average Lower Limb Length (mm)
E
Fig. 4—Bland-Altman analysis of limb length measurements. A–E, Graphs show limb length measurements for all patients for supine CT scout views versus upright full-length radiography (A), supine CT scout views versus biplanar linear radiography (B), upright full-length radiography versus biplanar linear radiography (C), supine CT scout views versus biplanar linear radiography composed leg method (D), and supine CT scout views versus biplanar linear radiography composed leg method (E). Note constant range of measurement differences with increasing average and consistent variability across all graphs. Detailed values are provided in Table 2.
surement method is able to compensate for this error due to an extension deficit in the knee joint, because the individual lengths of femur and tibia of a limb are taken into account and considered for calculation of the true limb length. Consequently, the lower limb length measurements using this method do not significantly differ from other modalities, and the subtle differences observed in the biplanar linear radiography standard measurements disappear. Usually, in the setting of planned total knee replacement, the composed leg measurement method is favored because it provides a more accurate estimation of postoperative lower limb length. Interreader agreement was very high using this meth-
od, with a mean difference between readers of only 0.5 mm. Mean alignment angles on biplanar linear radiography are equally calculated by both techniques and were slightly, but significantly, larger than those measured on supine CT scans and weight-bearing upright full-length radiography (average difference, 1.1° and 0.9°). The difference was even larger in patients with varus alignment angles greater than 10° (2.5° and 1.7°, respectively). Only in patients with a valgus deformity less than −10° alignment angles on CT scans were the differences larger than on biplanar linear radiography (2.6°), but because of the small number of patients in this group (n = 4), this finding has
0
Upper Limit (95% CI) Bias
0 –10 Lower Limit (95% CI)
–20 10 20 –20 –10 0 Average Lower Limb Angle (°)
A
10 0 –10
20 Upper Limit (95% CI) Bias Lower Limit (95% CI)
Difference (°)
10
to be interpreted with caution. Larger alignment angles in biplanar linear radiography indicate that malrotation of the lower limb may lead to distortion, either in varus or valgus alignment. Maximal and true limb alignment are measured in the frontal plane only in relation to each limb. This measurement plane can be reached only by compensating malrotation of the lower limb, using 3D models based on biplanar linear radiography. These angle measurements may thus be considered as the closest approximation to the true anatomy of the lower limb. According to Fang and Ritter [2], postoperative limb alignment angles in neutral or slight valgus alignment, optimally 2.4°–7.2° valgus, are considered to be optimal
20 Difference (°)
20 Difference (°)
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Guggenberger et al.
–20 –20 –10 0 10 20 Average Lower Limb Angle (°)
B
10 0 –10
Upper Limit (95% CI) Bias Lower Limit (95% CI)
–20 –20 –10 0 10 20 Average Lower Limb Angle (°)
C
Fig. 5—Bland-Altman analysis of limb alignment measurements. A–C, Graphs show limb alignment measurements for all patients for supine CT scout views versus upright full-length radiography (A), supine CT scout views versus biplanar linear radiography (B), and upright full-length radiography versus biplanar linear radiography (C). Note constant range of measurement differences with increasing average and consistent variability across all graphs. For detailed values refer to Table 2.
W166
AJR:202, February 2014
Assessment of Lower Limb Length and Alignment Using Different Modalities TABLE 3: Bland-Altman Analyses of Measurement Differences of Both Readers in Different Modalities
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Limb Length (mm)
Parameter Mean difference ± SD 95% CI Minimum, maximum Range
Limb Alignment Angle (°)
Supine CT
Upright Full-Length Radiographs
Upright Biplanar Linear Radiographs, Standard
Upright Biplanar Linear Radiographs, Composed
1.3 ± 7.0
−1.2 ± 6.6
−0.6 ± 2.8
−0.5 ± 3.4
−0.1 ± 1.4
−0.2 ± 1.3
0.2 ± 1.2
−0.1 to 2.7
−2.5 to 0.1
−1.1 to 0.0
−1.1 to 0.2
−0.4 to 1.5
−0.4 to 0.1
−0.1 to 0.4
−18, 45
−24, 12
−16, 4
−4, 4
−5, 7
−5, 5
−10, 9
63
36
20
8
12
10
19
Supine CT
Upright Full-Length Radiographs
Upright Biplanar Linear Radiographs
Note—All modalities showed no significant differences between both readers (p > 0.05).
to ensure long durability of the endoprosthesis. The measured mean angle differences between both readers and the different modalities are well within this recommended angle range and may thus be considered not clinically relevant. We are aware of certain limitations of this study. First, patients have to be able to stand upright in biplanar linear radiography. However, upright imaging also applies to upright full-length radiography, and, in our experience, many patients manage to stand up for a few seconds even if they are dependent on a wheelchair. Second, despite focusing on patients with severe osteoarthritis of the knee, there were not many patients with a significant extension deficit in the knee joint or severe varus or valgus deformity, potentially obscuring a significant advantage of 3D models. Finally, we did not compare radiation dose or time efficiency of the different modalities. Although a thorough evaluation of radiation dose lies beyond the scope of this work, this topic may be crucial in situations of repeated imaging, such as congenital deformities, or preoperative and repeated postoperative follow-up examinations in patients with complicated endoprosthesis surgery for severe osteoarthritis. In conclusion, measurements on 3D models based on upright biplanar linear radiographs allow lower limb length and alignment measure-
ments that are interchangeable with supine CT scans and upright full-length radiographs but with superior interreader agreement. References 1. Lang JE, Scott RD, Lonner JH, Bono JV, Hunter DJ, Li L. Magnitude of limb lengthening after primary total knee arthroplasty. J Arthroplasty 2012; 27:341–346 2. Fang D, Ritter MA. Malalignment: forewarned is forearmed. Orthopedics 2009; 32:pii 3. Bonner TJ, Eardley WG, Patterson P, Gregg PJ. The effect of post-operative mechanical axis alignment on the survival of primary total knee replacements after a follow-up of 15 years. J Bone Joint Surg Br 2011; 93-B:1217–1222 4. Parratte S, Pagnano MW, Trousdale RT, Berry DJ. Effect of postoperative mechanical axis alignment on the fifteen-year survival of modern, cemented total knee replacements. J Bone Joint Surg Am 2010; 92:2143–2149 5. Kawakami H, Sugano N, Yonenobu K, et al. Effects of rotation on measurement of lower limb alignment for knee osteotomy. J Orthop Res 2004; 22:1248–1253 6. Radtke K, Becher C, Noll Y, Ostermeier S. Effect of limb rotation on radiographic alignment in total knee arthroplasties. Arch Orthop Trauma Surg 2010; 130:451–457 7. Lonner JH, Laird MT, Stuchin SA. Effect of rotation and knee flexion on radiographic alignment
in total knee arthroplasties. Clin Orthop Relat Res 1996; 102–106 8. Myles CM, Rowe PJ, Walker CR, Nutton RW. Knee joint functional range of movement prior to and following total knee arthroplasty measured using flexible electrogoniometry. Gait Posture 2002; 16:46–54 9. Gheno R, Nectoux E, Herbaux B, et al. Three-dimensional measurements of the lower extremity in children and adolescents using a low-dose biplanar X-ray device. Eur Radiol 2012; 22:765–771 10. Thelen P, Delin C, Folinais D, Radier C. Evaluation of a new low-dose biplanar system to assess lower-limb alignment in 3D: a phantom study. Skeletal Radiol 2012; 41:1287–1293 11. Guenoun B, Zadegan F, Aim F, Hannouche D, Nizard R. Reliability of a new method for lowerextremity measurements based on stereoradiographic three-dimensional reconstruction. Orthop Traumatol Surg Res 2012; 98:506–513 12. Sabharwal S, Zhao C, McKeon JJ, McClemens E, Edgar M, Behrens F. Computed radiographic measurement of limb-length discrepancy: full-length standing anteroposterior radiograph compared with scanogram. J Bone Joint Surg Am 2006; 88:2243–2251 13. Janssen MM, Drevelle X, Humbert L, Skalli W, Castelein RM. Differences in male and female spino-pelvic alignment in asymptomatic young adults: a three-dimensional analysis using upright lowdose digital biplanar X-rays. Spine (Phila Pa 1976) 2009; 34:E826–E832
AJR:202, February 2014 W167
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Musculoskeletal Imaging Original Research
Assessment of Lower Limb Length and Alignment by Biplanar Linear Radiography: Comparison With Supine CT and Upright Full-Length Radiography Roman Guggenberger 1 Christian W. A. Pfirrmann1 Peter P. Koch2 Florian M. Buck1 Guggenberger R, Pfirrmann CWA, Koch PP, Buck FM
Keywords: biplanar linear x-ray scanner, limb alignment, limb length, supine CT, upright full-length radiography DOI:10.2214/AJR.13.10782 Received February 17, 2013; accepted after revision April 24, 2013. 1 Department of Diagnostic and Interventional Radiology, University Hospital Zurich, Raemistrasse 100, Zurich 8091, Switzerland. Address correspondence to R. Guggenberger ([email protected]). 2 Department of Orthopedic Surgery, Balgrist University Hospital, Zurich, Switzerland.
WEB This is a web exclusive article. AJR 2014; 202:W161–W167 0361–803X/14/2022–W161 © American Roentgen Ray Society
OBJECTIVE. The purpose of this article is to compare lower limb length and alignment measurements on supine CT, upright full-length radiography, and 3D models based on upright biplanar linear radiography. SUBJECTS AND METHODS. This study involved 51 consecutive patients (22 men and 29 women; mean age, 68.8 years; range, 43–92 years) who were scheduled for total knee replacement. Lower limb length and alignment angle were measured on CT, upright fulllength radiography, and 3D models based on biplanar linear radiography with standard and composed leg methods by two independent readers. Descriptive statistics of each modality were calculated. Measurements of different modalities were compared by paired Student t tests. Agreement between readers and modalities was assessed by Bland-Altman analyses. RESULTS. Mean (± SD) limb lengths were 783 ± 56.1 mm (range, 639–927 mm), 785 ± 53.0 mm (range, 655–924 mm), 780 ± 55.4 mm (range, 633–921 mm), and 783 ± 55.9 mm (range, 636–924 mm) for CT, upright full-length radiography, and 3D models based on biplanar linear radiography standard and composed leg measurements, respectively. Mean alignment angles were 2.3° ± 5.5° (range, −12° to 20°) for CT, 2.5° ± 6.7° (range, −17° to 18°) for upright full-length radiography, and 3.4° ± 6.6° (range, −14° to 18°) for 3D models based on biplanar linear radiography. No significant differences among modalities for mean limb length were found when using composed leg measurements in biplanar linear radiography. Very small but significant mean differences in angle measurements were seen for CT (−1.1° ± 2.5) and upright full-length radiography (−0.9° ± 3.1) compared with biplanar linear radiography. Bland-Altman analyses showed no significant differences between readers, with the highest agreement for biplanar linear radiography length measurements. CONCLUSION. Measurements on 3D models based on upright biplanar linear radiographs allow lower limb length and alignment angle measurements that are interchangeable with supine CT scans and upright full-length radiographs but with superior interreader agreement.
D
ifferences in limb length should be reliably quantified before knee replacement surgery, allowing intraoperative correction and symmetric postoperative length of the lower limbs [1]. In addition, a physiologic alignment should be restored. Postoperative varus or valgus deformity of more than 3° has been shown to markedly reduce the durability of total knee replacements and has been associated with a higher rate of surgical revision procedures [2]. Despite some controversy in the literature, a neutral postoperative limb alignment seems to be beneficial for long-term outcome after total knee prosthesis implantation [3, 4]. Today, several imaging modalities allow length and alignment measurements of the lower limb in either a supine or upright weight-
bearing position. The scout view of a supine CT scan may serve as a full-length radiograph but can be acquired only in a non-weight-bearing position. Conventional upright full-length radiography allows weight-bearing upright imaging of the patient but, analogous to supine CT scans, may be associated with measurement errors in cases of rotational misplacement or malpositioning of the lower extremities at the time of image acquisition [5–7]. To assess reliably length and alignment angle of the lower limb, full extension of the knee joint is required. This, however, may be impossible in cases of severe osteoarthritis or contractures of the extensor muscles [8]. Upright biplanar linear radiography with generation of 3D models based on simultaneously acquired frontal and lateral projections
AJR:202, February 2014 W161
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Guggenberger et al.
Fig. 1—52-year-old woman with severe osteoarthritis of knee. For all modalities (supine CT scans, upright full-length radiographs, and 3D models based on biplanar linear radiography), lower limb length was measured between center of femoral head and center of superior contour of talar trochlea (left leg). Alignment angle was defined by line through center of femoral head and intercondylar notch of femur and line between intercondylar eminence of tibia and center of ankle joint (right leg). Positive alignment values signified varus alignment, and negative values signified valgus alignment.
allows both upright imaging under weightbearing conditions and the depiction of 3D structures with consecutive compensation of extension deficits in the knee joints or rotation or projection differences occurring in mere 2D projection modalities. Measurements of the lower limb from 3D models based on biplanar linear radiography have already been shown to be comparable to 3D measurements based on CT in pediatric patients [9] and have shown increased accuracy for hip-knee-ankle angle measurements compared with 2D in ex vivo and in vivo conditions [10, 11]. Measurements from 3D models of the legs allow measurement of the limb length in 3D with and without taking account of an extension defi-
W162
Fig. 2—52-year-old woman with severe osteoarthritis of knee. A–C, Supine CT scout view (A), conventional upright full-length radiograph (B), and upright biplanar linear radiograph (C) of lower limbs are shown. Note slight pelvic rotation to right (A) and stitching artifacts from postprocessing (slight dark horizontal bands, B) whereas homogeneous depiction quality is perceived in panels A and C. Upright linear radiographs from anteroposterior (C) and lateral (not shown) projections were used for generation of 3D models. Measured limb length and alignment angles of both legs in this patient ranged between 721 and 738 mm and 0°–2° valgus for different modalities.
cit in the knee joint. Thus, postoperative limb length after implantation of total knee prostheses, which restores full extension in the knee joint, can be estimated. Biplanar linear radiography has therefore gained increasing attention in recent years in both the orthopedic and radiologic communities and may potentially replace traditional modalities for the assessment of lower limb geometry (i.e., length and alignment angle). However, the question arises whether limb length and alignment measurements using 3D models from upright biplanar linear radiographs are interchangeable with measurements on supine CT scans and weight-bearing upright full-length radiographs and whether measurement reliability is superior in patients with severe osteoarthritis of the knee joint. At present, to our knowledge, no comparison has been made between 3D models from upright
biplanar linear radiography and projectional 2D modalities in the supine (CT scan) or weight-bearing (upright full-length radiography) position under in vivo conditions. Thus, the purpose of our study was to compare lower limb length and alignment angle measurements using supine CT scout views, weight-bearing upright full-length radiography, and 3D models based on biplanar linear radiography. Subjects and Methods Institutional review board approval was obtained before the start of the study. In total, 59 patients with severe osteoarthritis of the knee scheduled for a total knee replacement were prospectively included after giving their informed consent. Eight patients had to be excluded from data analysis because of missing scaling items or insufficient image quality of upright full-length radiographs. Patients younger than 18
AJR:202, February 2014
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Assessment of Lower Limb Length and Alignment Using Different Modalities years, patients with Legg-Calvé-Perthes or developmental dysplasia of the hip, and patients who were unable to stand in an upright position would have been excluded also. Eventually, 51 patients (22 men and 29 women; mean age, 68.8 years; range, 43–92 years) were included in the final data analysis. As a routine preoperative imaging workup, all patients were scheduled for a CT scan of the hip joints, knee joints, and ankle joints for the production of patient-matched cutting blocks for knee prosthesis implantation surgery. The scout views of these CT scans were used for this study. In addition, an upright full-length radiography and a biplanar linear radiography were performed the same day.
CT Scans CT scans were performed using a 64-MDCT unit (Brilliance 64, Philips Healthcare). The patient was scanned in the supine position with the feet entering the gantry first. Scans included the entire lower limb from the iliac crests to the soles of the feet. Both legs were composed as much as possible, and care was taken to stabilize them in a neutral position without any rotation. Ankle joints were flexed in 90°, and the toes were pointed toward the ceiling. The anteroposterior scout view was acquired using the following settings: scan length, 120 cm; tube voltage, 80 kV; tube current, 300 mA; FOV, 500 mm; and matrix, 128 × 512.
Fig. 3—52-year-old woman with severe osteoarthritis of knee. A and B, Semiautomated generation of 3D model is based on matching of 3D model of femur and tibia with osseous contours of lower limb on frontal and lateral linear radiographs (A). Software requires identification of predefined osseous landmarks (e.g., femoral head, femoral condyles, tibial head, and medial and lateral malleolus) to generate 3D model. This model is then fine-tuned manually. Eventually, 3D limb lengths and limb alignments angle are calculated automatically by software (B).
A
Upright Full-Length Radiography Upright full-length radiography was performed on a fully digital radiography system (Ysio, Siemens Healthcare). Different preset tube voltages were chosen by the radiology technician according to the patient’s stature and weight (small, 75– 85 kV; medium, 77–90 kV; large, 77–96 kV), and tube currents were set according to automatic exposure measurements. A scaling item (a small radiopaque sphere with a diameter of 2.5 cm) was positioned next to the right lower leg before image acquisitions and was used for image calibration.
Biplanar Linear Radiography Anteroposterior and lateral linear radiographs were simultaneously acquired using a biplanar radiography system (EOS, EOS Imaging). The patient was centered in the upright position on the scanner platform. The scan included the whole lower limb from the hip joint to the ankle joint. Tube voltage of the x-ray tubes was chosen by the radiology technician among three different presets, according to the patients’ weight and stature (small, medium, and large; 80–104 kV with 200– 320 mA for anteroposterior projection and 100– 120 kV with 200–320 mA for lateral projection). A translational speed of the tubes of 5 was chosen from the vendor-specific scale, ranging from 1
B AJR:202, February 2014 W163
Guggenberger et al.
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
(fast) to 8 (slow). After scanning, the images were sent to an EOS workstation for secondary reconstruction and generation of 3D models.
Postprocessing and Measurements CT scout view and full-length radiography—The measurements were performed by two independent musculoskeletal radiologists with 3 and 7 years of experience. Lower limb length was defined as the distance from the center of the femoral head to the center of the superior contour of the talar trochlea. Lower limb alignment angle was defined by a line through the center of the femoral head and the intercondylar notch of the femur and a line between the intercondylar eminence of the tibia and the center of the ankle joint [12] (Figs. 1 and 2). Biplanar linear radiography—Biplanar linear radiographs were postprocessed using a vendorspecific workstation (sterEOS, EOS Imaging) by two independent radiology technicians with special training in 3D reconstruction based on biplanar linear radiography. The user of the reconstruction software manually marks predefined osseous landmarks, such as the femoral head, femoral condyles, tibial head, and medial and lateral malleolus. Then the software semiautomatically matches a 3D model of a femur and a tibia to the contour of femur and tibia on the frontal and lateral radiographs. On the basis of this 3D model, measurements of limb length and limb alignment are automatically calculated (Fig. 3). Limb length is calculated as the distance between the center of the femoral head and the center of the superior contour of the talar trochlea in 3D models (standard limb length measurement). In addition, there is another way to measure limb length, with the software taking into account extension deficits, where the limb length is calculated by adding up the 3D lengths of the femur and tibia. We call this measurement method the “composed leg” limb length measurement. Limb alignment angles, however, remain unchanged.
Statistical Analysis All calculations were performed using commercially available software (SPSS Statistics, version 19, IBM). Descriptive statistics (mean [± SD] and range)
were calculated for lower limb length and alignment of the three different modalities and both readers. Interreader agreement and agreements between modalities were assessed by Bland-Altman analysis and paired two-tailed Student t testing. Bland-Altman plots were generated to graphically depict scatter of measurement differences. A p value less than 0.05 was considered statistically significant.
Results Lower Limb Length Mean limb lengths were 783 ± 56.1 mm (range, 639–927 mm) for CT scout views, 785 ± 53.0 mm (range, 655–924 mm) for upright full-length radiographs, and 780 ± 55.4 mm (range, 633–921 mm) for 3D models based on biplanar linear radiographs (Table 1). Composed leg limb length measurements revealed a mean limb length of 783 ± 55.9 mm (range, 636–924 mm). Small statistically significant (p < 0.001) differences were seen among the different modalities for lower limb length measurements. Limb lengths measured using 3D models based on biplanar linear radiographs were significantly shorter than measurements on CT scans and upright full-length radiographs. No differences were seen between limb length measurements on CT scans and upright full-length radiographs (p = 0.057). On average, limb length measurements using 3D models based on biplanar linear radiographs were 2.7 ± 5.5 mm shorter than measurements on CT scans and 5.4 ± 13.1 mm shorter than measurements performed on upright full-length radiographs. Regarding the composed leg measurement method, no statistically significant differences were seen in comparison with CT scans and biplanar linear radiographs (Table 2). Lower Limb Alignment Mean alignment angles were 2.3° ± 5.5° (range, −12° to 20°) for CT scout views, 2.5° ± 6.7° (range, −17° to 18°) for upright full-length radiographs, and 3.4° ± 6.6° (range, −14° to 18°) for 3D models based on biplanar linear radiographs (Table 1). Small statistically signifi-
cant (p < 0.001) differences were seen among the different modalities for angle measurements. Compared with measurements using 3D models based on biplanar linear radiographs, angles measured on CT scans were 1.1° ± 2.5° smaller and measurements on upright fulllength radiographs were 0.9° ± 3.1° smaller. No differences were seen for alignment measurements between CT scans and upright fulllength radiographs (p = 0.504) (Table 2). Intermodality Agreement Bland-Altman analyses for comparisons of different modalities showed a constant range of measurement differences and scatter with increasing average for both length and alignment angle measurements. Consistent variability of measurements across all graphs was seen. Further results concerning specific patient subgroups (varus alignment, valgus alignment, varus angle > 10°, and valgus angle > 10° as measured on upright fulllength radiography) are shown in Table 2 and Figures 4 and 5. Interreader Agreement Interreader agreement as calculated with Bland-Altman analyses was high and showed no significant differences between both readers in all modalities. The smallest mean lower limb length difference and range between both readers were seen in measurements on 3D models based on biplanar linear radiographs using the composed leg measurement (mean difference, −0.5 ± 3.4 mm; range, 8 mm) (Table 3). Discussion Despite small statistically significant differences, we found overall clinically comparable lower limb length and alignment angle measurements on supine CT scans, weight-bearing upright full-length radiographs, and biplanar linear radiographs in this study. Although patient posture seems to not significantly affect both measures, Bland-Altman analyses detected small but significant differences between 3D models from weight-bearing bipla-
TABLE 1: Descriptive Statistics of Mean Limb Length and Alignment Measurements for Different Imaging Modalities and Both Readers Reader 1 Imaging Modality
Reader 2
Limb Length (mm)
Limb Alignment Angle (°)
Limb Length (mm)
Limb Alignment Angle (°)
Supine CT scout view
783 ± 55.8
2.3 ± 5.5
782 ± 56.7
2.4 ± 5.5
Upright full-length radiographs
785 ± 53.6
2.5 ± 6.9
786 ± 52.7
2.6 ± 6.7
Upright biplanar linear radiographs
779 ± 55.3
3.5 ± 6.7
781 ± 55.7
3.4 ± 6.6
Composed limb length measurement
783 ± 55.8
784 ± 56.0
Note—Data are mean ± SD.
W164
AJR:202, February 2014
Note—Bold type indicates a statistically significant difference (p < 0.05).
−3, 1
4 14
−3, 11 −5, 3
8 26
−8, 18 −8, 18
26 5
1, 4 −6, 2
8 7
−2, 5 −7, 8
15 14
−7, 8 2, 7
9 14
−12, 2 −1, 8
8 22
−18, 5 −18, 8
25 Range
Minimum, maximum
−0.9 ± 3.1 −0.6 ± 3.7 −1.6 ± 1.6 −1.7 ± 3.7 −1.1 ± 1.8
−1.5 to −0.3 −1.5 to 0.3 −2.1 to −1.0 −0.6 to 4.1 −3.9 to 1.6 0.9–4.3
2.6 ± 1.1 −2.5 ± 2.2 0.3 ± 1.9 −1.9 ± 2.4
−1.6 to −0.6 −2.5 to −1.3 −0.3 to 1.0 −3.9 to −1.0 0–7.6
−4.2 ± 3.6 −3.8 ± 2.4 −1.1 ± 2.5
−6.4 to −2 95% CI
1.9 ± 1.9 −0.2 ± 3.4 −1.3 ± 3.5
−0.9 to 0.4 −2 to −0.4
Mean difference ± SD
Composed leg length
Lower limb alignment angle (°)
1.2–2.6
27 47 43 83 83 12 18 35 18 35
−9, 18
25 38 32 87 87 15 23 35 39 40 33 60 37 81 81 Range
−5, 20 −19, 19
−30, 17 −15, 28
−11, 22 −52, 36
−55, 28 −55, 28
−52, 36 1, 14
−2, 10 −10, 8
−4, 19 −21, 14
−26, 10 −10, 8
−20, 19 −21, 19
−26, 10
−15, 18 −22, 38 −19, 18 −31, 51 −31, 51 Composed leg length
Minimum, maximum
−7 to 10 −7.4 to 24 1.7–7.3 2.2–9.5 2.8–7.9 −3 to 15 −0.1 to 8.6 0.1–4.5 1.7–4.1 1.6–3.8
−1.5 to 0.2 −3.4 to 1.6 −3.7 to 2.5 −5.1 to 10.6 −0.8 to 4.6 −1.5 to 6.0 −2.0 to 4.6 −13 to 7 −12 to 23 Composed leg length
95% CI
−5.4 to 0.1 −6.7 to 0.9 −5.6 to 1.2 −9.3 to 15 −25 to 20
−1.7 to 0.3
−3.3 ± 16 5.4 ± 11
1.6 ± 13 −8.4 ± 10 4.5 ± 8
1.3 ± 9 2.3 ± 16
5.8 ± 15 5.4 ± 13
1.9 ± 14 2.8 ± 4.9
5.8 ± 5.5 4.4 ± 6.7 2.3 ± 6.3
−0.9 ± 7.3 −0.6 ± 4.9 −0.6 ± 3.7
2.9 ± 5.1 2.7 ± 5.5 −2.7 ± 13.9 −2.9 ± 15.6 −2.2 ± 9.8 −2.8 ± 18.9 −2.6 ± 14
Composed leg length
Mean difference ± SD
Lower limb length (mm)
Limb, Measurement
−0.7 ± 5.1
Valgus > 10° (n = 4) Varus > 10° (n = 12) Valgus (n = 34) Varus (n = 68) All (n = 102) Valgus > 10° (n = 4) Varus > 10° (n = 12) Valgus (n = 34) Varus (n = 68) All (n = 102) Valgus > 10° (n = 4) Varus > 10° (n = 12) Valgus (n = 34) Varus (n = 68) All (n = 102)
Upright Full-Length Radiographs vs Upright Biplanar Linear Radiographs Supine CT vs Upright Biplanar Linear Radiographs Supine CT vs Upright Full-Length Radiographs
TABLE 2: Bland-Altman Analyses of Measurement Differences in Different Modalities for Limb Length and Alignment Measurements in All Patients and Subanalyses for Patients With Varus, Valgus, Varus > 10°, and Valgus > 10°
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Assessment of Lower Limb Length and Alignment Using Different Modalities nar linear radiography and both other modalities. However, these differences were in the order of magnitude of the measurement accuracies themselves. Interreader agreement was high for all modalities, with the highest agreement seen in lower limb length measurements using 3D models based on biplanar linear radiographs. Upright full-length radiography offers fast depiction of the lower limb in an upright patient position. Technically, it consists of three projections at different angulations that are stitched together and visually corrected by the technician to form a single image. Hence, upright full-length radiographs may, in addition to the impact of posture, be associated with a certain distortion effect when compared with supine CT scans, leading to the observed differences in severe misalignment angles [5–7]. Similarly, low-dose scout views in CT scans are also a projection of 3D structures onto a 2D plane but performed with the patient in a non-weight-bearing supine position. In contrast to upright full-length radiography, CT scout views are not associated with distortion artifacts due to possible stitching errors. Therefore, observed mean differences between both modalities for lower limb length and alignment angles were small (average, 2.7 mm and 0.2°, respectively) and may be considered clinically irrelevant [1, 2]. Interestingly, larger differences of limb alignment angle measurements were seen when comparing supine CT scans to weight-bearing upright full-length radiography in patients with varus or valgus alignment of more than 10° (average, −4.2° for varus angles > 10° and −3.8° for valgus angles < −10°). This increase in difference between both modalities may be caused by increased lower limb misalignment angles under upright weight-bearing conditions during upright full-length radiography imaging. Therefore, marked varus or valgus deformity posture during imaging seems to increasingly affect measured lower limb alignment angles. Upright biplanar linear radiography with reconstruction of 3D models based on simultaneously acquired contiguous frontal and lateral projections combines the advantage of a physiologic upright weight-bearing position during imaging with the possibility of 3D model generation at a low radiation dose [9, 13]. To be able to generate exact 3D models, the patient needs to be positioned in a small step position to allow delineation of the posterior contour of the femoral condyles of both legs on the lateral view. Overall, limb length measurements were comparable among different modalities, and mean differences were small from a clinical point of view. Limb length measurements on 3D models using standard measurements were, however, significantly smaller compared with supine CT scans and weight-bearing upright full-length radiography. These mean differences increased in patients with severe varus and valgus deformity (average, 4.4 and 5.8 mm, respectively) comparing biplanar linear radiography to CT. Limb length is measured on 3D models as the distance from the center of the femoral head to the center of the upper ankle joint. Thus, flexion of the knee joint at the time of image acquisition leads to a falsely short limb length measurement in these 3D models. The composed leg mea-
AJR:202, February 2014 W165
0 Bias
–20
Lower Limit (95% CI)
Upper Limit (95% CI) Bias
0
Lower Limit (95% CI)
–20 600 700 800 900 1000 Average Lower Limb Length (mm)
Upper Limit (95% CI)
20
Bias
0 –20
Lower Limit (95% CI)
600 700 800 900 1000 Average Lower Limb Length (mm)
C
20
Upper Limit (95% CI)
0
Bias Lower Limit (95% CI)
–20
Difference (mm)
B
Difference (mm)
A
20
Difference (mm)
Upper Limit (95% CI)
Difference (mm)
Difference (mm)
20
600 700 800 900 1000 Average Lower Limb Length (mm)
600 700 800 900 1000 Average Lower Limb Length (mm)
D
Upper Limit (95% CI)
20 0
Bias
–20
Lower Limit (95% CI)
600 700 800 900 1000 Average Lower Limb Length (mm)
E
Fig. 4—Bland-Altman analysis of limb length measurements. A–E, Graphs show limb length measurements for all patients for supine CT scout views versus upright full-length radiography (A), supine CT scout views versus biplanar linear radiography (B), upright full-length radiography versus biplanar linear radiography (C), supine CT scout views versus biplanar linear radiography composed leg method (D), and supine CT scout views versus biplanar linear radiography composed leg method (E). Note constant range of measurement differences with increasing average and consistent variability across all graphs. Detailed values are provided in Table 2.
surement method is able to compensate for this error due to an extension deficit in the knee joint, because the individual lengths of femur and tibia of a limb are taken into account and considered for calculation of the true limb length. Consequently, the lower limb length measurements using this method do not significantly differ from other modalities, and the subtle differences observed in the biplanar linear radiography standard measurements disappear. Usually, in the setting of planned total knee replacement, the composed leg measurement method is favored because it provides a more accurate estimation of postoperative lower limb length. Interreader agreement was very high using this meth-
od, with a mean difference between readers of only 0.5 mm. Mean alignment angles on biplanar linear radiography are equally calculated by both techniques and were slightly, but significantly, larger than those measured on supine CT scans and weight-bearing upright full-length radiography (average difference, 1.1° and 0.9°). The difference was even larger in patients with varus alignment angles greater than 10° (2.5° and 1.7°, respectively). Only in patients with a valgus deformity less than −10° alignment angles on CT scans were the differences larger than on biplanar linear radiography (2.6°), but because of the small number of patients in this group (n = 4), this finding has
0
Upper Limit (95% CI) Bias
0 –10 Lower Limit (95% CI)
–20 10 20 –20 –10 0 Average Lower Limb Angle (°)
A
10 0 –10
20 Upper Limit (95% CI) Bias Lower Limit (95% CI)
Difference (°)
10
to be interpreted with caution. Larger alignment angles in biplanar linear radiography indicate that malrotation of the lower limb may lead to distortion, either in varus or valgus alignment. Maximal and true limb alignment are measured in the frontal plane only in relation to each limb. This measurement plane can be reached only by compensating malrotation of the lower limb, using 3D models based on biplanar linear radiography. These angle measurements may thus be considered as the closest approximation to the true anatomy of the lower limb. According to Fang and Ritter [2], postoperative limb alignment angles in neutral or slight valgus alignment, optimally 2.4°–7.2° valgus, are considered to be optimal
20 Difference (°)
20 Difference (°)
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Guggenberger et al.
–20 –20 –10 0 10 20 Average Lower Limb Angle (°)
B
10 0 –10
Upper Limit (95% CI) Bias Lower Limit (95% CI)
–20 –20 –10 0 10 20 Average Lower Limb Angle (°)
C
Fig. 5—Bland-Altman analysis of limb alignment measurements. A–C, Graphs show limb alignment measurements for all patients for supine CT scout views versus upright full-length radiography (A), supine CT scout views versus biplanar linear radiography (B), and upright full-length radiography versus biplanar linear radiography (C). Note constant range of measurement differences with increasing average and consistent variability across all graphs. For detailed values refer to Table 2.
W166
AJR:202, February 2014
Assessment of Lower Limb Length and Alignment Using Different Modalities TABLE 3: Bland-Altman Analyses of Measurement Differences of Both Readers in Different Modalities
Downloaded from www.ajronline.org by Cornell University Library on 06/09/14 from IP address 132.236.27.217. Copyright ARRS. For personal use only; all rights reserved
Limb Length (mm)
Parameter Mean difference ± SD 95% CI Minimum, maximum Range
Limb Alignment Angle (°)
Supine CT
Upright Full-Length Radiographs
Upright Biplanar Linear Radiographs, Standard
Upright Biplanar Linear Radiographs, Composed
1.3 ± 7.0
−1.2 ± 6.6
−0.6 ± 2.8
−0.5 ± 3.4
−0.1 ± 1.4
−0.2 ± 1.3
0.2 ± 1.2
−0.1 to 2.7
−2.5 to 0.1
−1.1 to 0.0
−1.1 to 0.2
−0.4 to 1.5
−0.4 to 0.1
−0.1 to 0.4
−18, 45
−24, 12
−16, 4
−4, 4
−5, 7
−5, 5
−10, 9
63
36
20
8
12
10
19
Supine CT
Upright Full-Length Radiographs
Upright Biplanar Linear Radiographs
Note—All modalities showed no significant differences between both readers (p > 0.05).
to ensure long durability of the endoprosthesis. The measured mean angle differences between both readers and the different modalities are well within this recommended angle range and may thus be considered not clinically relevant. We are aware of certain limitations of this study. First, patients have to be able to stand upright in biplanar linear radiography. However, upright imaging also applies to upright full-length radiography, and, in our experience, many patients manage to stand up for a few seconds even if they are dependent on a wheelchair. Second, despite focusing on patients with severe osteoarthritis of the knee, there were not many patients with a significant extension deficit in the knee joint or severe varus or valgus deformity, potentially obscuring a significant advantage of 3D models. Finally, we did not compare radiation dose or time efficiency of the different modalities. Although a thorough evaluation of radiation dose lies beyond the scope of this work, this topic may be crucial in situations of repeated imaging, such as congenital deformities, or preoperative and repeated postoperative follow-up examinations in patients with complicated endoprosthesis surgery for severe osteoarthritis. In conclusion, measurements on 3D models based on upright biplanar linear radiographs allow lower limb length and alignment measure-
ments that are interchangeable with supine CT scans and upright full-length radiographs but with superior interreader agreement. References 1. Lang JE, Scott RD, Lonner JH, Bono JV, Hunter DJ, Li L. Magnitude of limb lengthening after primary total knee arthroplasty. J Arthroplasty 2012; 27:341–346 2. Fang D, Ritter MA. Malalignment: forewarned is forearmed. Orthopedics 2009; 32:pii 3. Bonner TJ, Eardley WG, Patterson P, Gregg PJ. The effect of post-operative mechanical axis alignment on the survival of primary total knee replacements after a follow-up of 15 years. J Bone Joint Surg Br 2011; 93-B:1217–1222 4. Parratte S, Pagnano MW, Trousdale RT, Berry DJ. Effect of postoperative mechanical axis alignment on the fifteen-year survival of modern, cemented total knee replacements. J Bone Joint Surg Am 2010; 92:2143–2149 5. Kawakami H, Sugano N, Yonenobu K, et al. Effects of rotation on measurement of lower limb alignment for knee osteotomy. J Orthop Res 2004; 22:1248–1253 6. Radtke K, Becher C, Noll Y, Ostermeier S. Effect of limb rotation on radiographic alignment in total knee arthroplasties. Arch Orthop Trauma Surg 2010; 130:451–457 7. Lonner JH, Laird MT, Stuchin SA. Effect of rotation and knee flexion on radiographic alignment
in total knee arthroplasties. Clin Orthop Relat Res 1996; 102–106 8. Myles CM, Rowe PJ, Walker CR, Nutton RW. Knee joint functional range of movement prior to and following total knee arthroplasty measured using flexible electrogoniometry. Gait Posture 2002; 16:46–54 9. Gheno R, Nectoux E, Herbaux B, et al. Three-dimensional measurements of the lower extremity in children and adolescents using a low-dose biplanar X-ray device. Eur Radiol 2012; 22:765–771 10. Thelen P, Delin C, Folinais D, Radier C. Evaluation of a new low-dose biplanar system to assess lower-limb alignment in 3D: a phantom study. Skeletal Radiol 2012; 41:1287–1293 11. Guenoun B, Zadegan F, Aim F, Hannouche D, Nizard R. Reliability of a new method for lowerextremity measurements based on stereoradiographic three-dimensional reconstruction. Orthop Traumatol Surg Res 2012; 98:506–513 12. Sabharwal S, Zhao C, McKeon JJ, McClemens E, Edgar M, Behrens F. Computed radiographic measurement of limb-length discrepancy: full-length standing anteroposterior radiograph compared with scanogram. J Bone Joint Surg Am 2006; 88:2243–2251 13. Janssen MM, Drevelle X, Humbert L, Skalli W, Castelein RM. Differences in male and female spino-pelvic alignment in asymptomatic young adults: a three-dimensional analysis using upright lowdose digital biplanar X-rays. Spine (Phila Pa 1976) 2009; 34:E826–E832
AJR:202, February 2014 W167
