- Email: [email protected]
Brief report: validation of a system for automated measurement of knee laxity
Clinical Biomechanics 19 (2004) 308–312 www.elsevier.com/locate/clinbiomech
Brief report: validation of a system for automated measurement of knee laxity M.T. Thompson a
a,*
, M.A. Conditt a, S.K. Ismaily a, A. Agarwal b, P.C. Noble
a,b
Institute of Orthopedic Research and Education, 6550 Fannin Street, Suite 2512, Houston, TX 77030, USA b Baylor College of Medicine, Houston, TX 77030, USA Received 24 September 2003; accepted 4 December 2003
Abstract Objective. To determine the accuracy and repeatability of an automated quantitative fluoroscopic imaging system for measuring knee laxity. Design. Cadaveric validation study. Background. Current methods of measuring anterior–posterior laxity lack sufficient accuracy and repeatability. A commercially developed fluoroscopic software package, capable of measuring laxity, required validation. Methods. Five human cadaveric knees were used. A constant force of 130 N was applied anteriorly and posteriorly in turn to the tibia of each knee with the femur fixed in 30 and 90 of flexion. Quantitative fluoroscopic measurements of anterior–posterior laxity were determined using image analysis software. Fluoroscopic results were compared to the true anterior–posterior displacements of the tibia, which were simultaneously recorded using linear transducers directly attached to the cadaveric specimens. Results. The quantitative fluoroscopic method underestimated laxity by an average of 0.40 mm with a root mean square error of 0.49 mm. The 95% confidence intervals for anterior and posterior laxity error were calculated to be )0.99 to 0.25 mm and )0.89 to 0.03 mm, respectively, where a negative error represents an underestimation. Conclusions. The quantitative fluoroscopic method offers a dramatic improvement in accuracy over current laxity measurement techniques and acceptable repeatability for assessing ligament damage. Relevance The considerably more accurate, validated measurement system of this study could improve ligament assessment and diagnosis, and the recognition of injuries otherwise undetected with current methods. 2004 Elsevier Ltd. All rights reserved. Keywords: Knee; Laxity; Fluoroscope; AP drawer
1. Introduction Anterior–posterior (AP) knee laxity is routinely assessed in evaluating knee ligament injuries and the success of treatment. Several mechanical, non-invasive methods have been developed to measure AP laxity using devices, such as arthrometers, attached to the lower leg. Although these devices are commonly used to obtain objective measurements of knee laxity, they have been shown to significantly overestimate true laxity and often
*
Corresponding author. E-mail address: [email protected] (M.T. Thompson).
0268-0033/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.clinbiomech.2003.12.004
exhibit poor repeatability (Uh et al., 2001; Fleming et al., 2002; Fleming et al., 1992; Daniel et al., 1985). The KT-1000 Knee Arthrometer (MedMetrics, San Diego, CA, USA) is the most widely used knee laxity measurement system (Daniel et al., 1985). When externally strapped to the lower leg, the KT-1000 records tibial translations relative to the patella during AP loading. A confidence interval of 2.4 mm within subjects and root mean square (RMS) errors as high as 3.9 mm have been calculated for the KT-1000 when compared to highly accurate (0.2 mm RMS error) Roentgen stereophotogrammetry analysis (RSA) (Fleming et al., 2002; Karrholm et al., 1988; Steiner et al., 1990; Wroble et al., 1990). RSA, however, is an invasive technique, requiring the implantation of tantalum beads into the femur and
M.T. Thompson et al. / Clinical Biomechanics 19 (2004) 308–312
tibia to provide radiographic landmarks imaged with biplanar radiography. Planar stress radiography is another common method to measure laxity using sequential lateral radiographs taken during anterior and posterior shear loads. The relative changes in displacement of the posterior aspect of the tibia are measured relative to the femoral condyles. The inaccuracy of planar stress radiography has been reported as high as 5.0 mm when compared to RSA (Fleming et al., 2002). Computer-assisted, fluoroscopic measurements offer non-invasive, real-time, objective quantification of AP laxity, but are seldom used clinically due to the lack of automated measurement software, poorly developed clinical guidelines that use knee laxity measurements, and concerns relating to radiation exposure. Moreover, though these techniques have been utilized in laboratory studies, they presently lack experimental validation for measurement of AP laxity. Studies using fluoroscopy to measure the complex 3D kinematics of total knee replacements report translational errors of 0.2–0.5 mm and out-of-plane errors as great as 6.0 mm (Stiehl et al., 1995; Banks and Hodge, 1996). Those studies, however, used accurately known 3D geometric properties of the implants or required computerized tomography (CT) of the bones to track in-plane and out-of-plane behavior, rather than tracking the in-plane motion of the femur and tibia directly from the fluoroscopic image. An accurate and reproducible method for measuring knee laxity that can be used in routine clinical practice would facilitate improved treatment guidelines for patients with knee ligament injuries. The accuracy of fluoroscopic measurements can be influenced by several factors including the extent of out-of-plane motion during the laxity exam, and the resolution and quality of the fluoroscope and its calibration equipment. The objective of this study was to determine the accuracy of two-dimensional, sagittal plane knee laxity measurements from a quantitative fluoroscopic imaging system through direct comparison with physical measurements in cadaveric knees.
309
2. Methods Five fresh frozen left cadaveric lower limbs (all males, ages 52–59) with no evidence of ligamentous pathology were cut approximately 30 cm above and below the joint line. The femur and tibia of each specimen was potted in PVC pipe and mounted in a two degree-of-freedom apparatus that allowed only anteroposterior motion of the tibia and superior/inferior motion of the femur (Fig. 1). The device was constructed completely of non-ferrous metal to minimize any distortion of the fluoroscopic image. The continuous AP motions of the femur and tibia were directly measured using two linear displacement transducers (UniMeasure, Inc., Corvallis, OR, USA), aligned in the AP direction and attached to the bones using radiolucent screws. No instrumentation was visible in the fluoroscopic image and no markers or beads were implanted in the bones (Fig. 1). The fluoroscopic imaging system (kimax 1024, Medical Metrics, Inc, Houston, TX, USA, camera resolution: 1024 · 1024 pixels, intensifier resolution: 0.302 mm/pixel) was positioned prior to testing so that the knee joint was centered on the fluoroscopic image. Alternating anterior and posterior forces of 132 N were applied to the tibia in 4-s cycles via a pneumatic cylinder. The knees were allowed to rest unloaded between runs. During each run, anteroposterior displacement of the femur and tibia were recorded at 30 Hz by the linear transducers placed parallel to the image intensifier and simultaneously by the image analysis software from Medical Metrics, Inc. Data collection was synchronized so that for each fluoroscopic frame there was a corresponding transducer value. The knee was loaded for one cycle initially to precondition the knee and obtain a repeatable starting and return point. Three runs with 3–5 cycles per run were performed with each knee held in 30 and 90 of flexion, for both anteriorly and posteriorly directed forces. One recorded fluoroscopic frame was chosen from each cycle of each run for comparison with the transducer measurements, generating a total of 275 laxity measurements. Prior to analysis of the fluoroscopic images, geometric distortion
Fig. 1. The experimental setup and corresponding fluoroscopic image.
310
M.T. Thompson et al. / Clinical Biomechanics 19 (2004) 308–312
transducer-based measurements of laxity of each knee, measured as one standard deviation about the mean, ranged from 0.06 to 1.78 mm for the five specimens. Of the two flexion angles and the two loading directions, the largest average error occurred during posterior loading at 30 (underestimation of 0.62 mm). The largest underestimation and overestimation of laxity were 0.97 and 0.53 mm, respectively. The 95% confidence intervals for anterior and posterior laxity error were calculated to be )0.99 to 0.25 mm and )0.89 to 0.03 mm, respectively, where a negative error represents an underestimation.
due to the image intensifier and the magnetic field of the environment was minimized using a distortion correction algorithm. The AP laxity of the knee was defined as the change in relative position of the tibia with respect to the femur in the anatomic AP direction, before and after loading. From the transducers, laxity was simply calculated by subtracting the femoral displacement from the tibial displacement. From the fluoroscopic images the outline of each bone was traced and anatomic landmarks in the first frame of the fluoroscopic imaging sequence were identified. The outer borders of the tibia and femur were traced based on a generic template provided by the software. The software tracked each bone through all frames of the fluoroscopic imaging sequence based on the unique radiographic appearance of the bone. New coordinates of the anatomic landmarks were calculated for each frame. The landmarks were chosen so that the AP translation of the tibia could be calculated relative to the femur. Any sensitivity of the tracking process to operator variation in tracing the outline of the bones would be minimized by the stabilization feature of the software which holds either the tibia or femur in a fixed position on the screen as the fluoroscopic imaging sequence is viewed. The accuracy of tracking was verified visually using this feature, and minor mistakes in the tracking process could be corrected by the user. The error in each automated laxity measurement was calculated by subtracting the fluoroscopic estimate of AP translation of the tibia with respect to the femur from the true value of AP laxity as recorded by the linear transducers. Paired and unpaired comparisons were performed between laxity, error, loading direction, and flexion angle. Correlation analyses were also performed between fluoroscopic and true measurements, and between laxity and measured error.
4. Discussion In this study, the accuracy of a novel method of measuring the AP laxity of the knee using a non-invasive quantitative fluoroscopic technique was evaluated. The nominal resolution of this system is approximately 0.3 mm per pixel. Larger errors might be expected with a lower resolution imaging system, although the software used to track motion in the fluoroscopic images interpolates between frames, and is thereby capable of giving subpixel accuracy. Both the transducers and the fluoroscopic method used in this study detected significant differences and similar trends in AP laxity between anterior and posterior loads and between 30 and 90 of flexion. On average, the fluoroscopic method underestimated the true laxity by 0.40 (SD, 0.28) mm with an RMS error of 0.49 mm. In laxity measurement, the 95% confidence interval for error is perhaps more relevant. The results have shown that regardless of loading direction, the magnitude of the 95% confidence interval for laxity error is less than 1.26 mm. There was little correlation ()0.314) between AP laxity and AP error, suggesting that fluoroscopic error does not increase with increased laxity. The source of the apparently systematic error was not clear, since both the fluoroscopic imaging system and the linear transducers were calibrated. The linear displacement transducers used to measure the true AP displacements have essentially infinite resolution with a repeatability of 0.03% full scale (0.02 mm) and a linearity of 1% full scale (0.50 mm). The data acquisition system limited the obtainable transducer resolution to approximately 0.01 mm. One possible source of the
3. Results The fluoroscopic measurement system slightly underestimated AP laxity for both anterior and posterior loading, and at both flexion angles of 30 and 90 (Table 1). The average error across all specimens and both loading directions was calculated to be an underestimation of 0.40 (SD, 0.28) mm. Variability in the
Table 1 Average values of AP laxity using direct transducer measurement and using the fluoroscopic system during anterior and posterior loading, and at 30 and 90 of flexion True laxity Fluoro laxity Error P value
Anterior
SD
Posterior
SD
P value
30
SD
90
SD
P value
3.00 2.62 )0.38 <0.01
1.23 1.23
3.33 2.90 )0.43 <0.01
1.36 1.27
0.03 0.05
3.46 2.92 )0.54
2.16 1.89
2.89 2.61 )0.28
1.03 1.12
<0.01 0.04
All values are in mm.
M.T. Thompson et al. / Clinical Biomechanics 19 (2004) 308–312
systematic error was the miscalculation of the magnification in the fluoroscopic images, which was estimated based on the distance of the midsagittal plane of the knee from the plane of the image intensifier, measured using a standard metric ruler. Clearly, if the systematic errors were removed, the error in the fluoroscopic measurements would be significantly less. A systematic, empirical-based correction could be applied to the estimated magnification to reduce subjective errors in the measurements. Since the soft tissue of the human knee can respond differently to even consecutive loads of the same magnitude, repeatability of AP laxity measurement using the fluoroscopic method could not be uniquely determined in this study. Perhaps more representative of quantitative fluoroscopic repeatability is one standard deviation of error in the laxity measurement (0.28 mm). This suggests that although the fluoroscopic method does underestimate laxity, its accuracy varies only slightly between measurements. There was no significant difference in the error of laxity measurements recorded in response to anterior and posterior loading, however, a significant difference (P < 0:001) was found between the 30 and 90 flexion angles ()0.54 vs. )0.27 mm, respectively). Previous work has assessed the accuracy of mechanical methods of measuring the AP laxity of the knee such as the KT-1000 arthrometer and the Vermont knee laxity device (VKLD) (Uh et al., 2001; Fleming et al., 2002; Steiner et al., 1990; Wroble et al., 1990). Fleming et al. (2002) performed AP laxity measurements on 15 subjects and compared the results obtained using the KT-1000, planar stress radiography, and RSA. Assuming RSA to be the true value, RMS errors of 5.0 and 3.9 mm were calculated for the KT-1000 and planar stress radiography, respectively. In another study, the repeatability of laxity measurements recorded by experienced operators applying a 130 N load was found to be 1.4 (SD, 1.4 mm) with the KT-1000 and as 2.4 (SD, 2.5 mm) with the VKLD (Uh et al., 2001). Quantitative fluoroscopic errors determined in previous studies examining total knee replacement kinematics (Stiehl et al., 1995; Banks and Hodge, 1996) are not relevant in fluoroscopic measurement of AP laxity. This study has several limitations. Because of the alignment and constraints of the testing device, the tibia was only free to move in the AP direction, thus, measurement of out of plane motion was unnecessary and the linear transducers were sufficient. In a clinical setting, however, out of plane motion may be unavoidable, and the errors due to out-of-plane motion remain to be characterized. The testing device also applied repeatable, controlled loads to the tibia. To achieve a clinical test for knee laxity that minimizes variability due to the individual performing the test, a mechanism to apply controlled loads to the tibia would be of benefit. Such a device is not readily available for clinical use. There are
311
also knee ligament, meniscal and cartilage injuries that may not be detectable with a two-dimensional test performed under fluoroscopic imaging. Finally, there was minimal rotation of the femur or tibia during the cadaver loading experiments. In clinical practice, sagittal plane rotations might occur, and knee laxity is poorly defined in the presence of combined AP translation and sagittal plane rotation. The clinical value of more accurate knee laxity measurements is not well-defined. Is it really advantageous to have knee laxity measurements that are accurate to less than 1 mm compared to measurements that are accurate to 3 or more mm? Perhaps with improved accuracy, it may be possible to detect a partial ligament injury, or detect a ligament injury when swelling or muscle action is limiting motion, or detect changes in laxity following surgery that may predict if the reconstruction is failing. The improved accuracy would at least facilitate laboratory and clinical experiments to address the above hypotheses. Perhaps more importantly, knee laxity measurements that are minimally effected by inter-tester variation and are free from errors due to soft-tissue artifacts could be used by general practitioners to determine which patients need to be seen by an orthopedic surgeon.
5. Conclusion The quantitative fluoroscopic technique evaluated in this study underestimated knee laxity by an average of 0.40 (SD, 0.28) mm for combined anterior and posterior directions, with a 95% confidence level of 1.09 mm and an RMS error of 0.49 mm. In comparison, the RMS errors of mechanical arthrometers utilized clinically is reported to be as high as 3.9 mm (Uh et al., 2001; Fleming et al., 2002; Daniel et al., 1985; Fleming et al., 2001). In view of these findings, the quantitative fluoroscopic method offers a dramatic improvement in accuracy over current AP laxity techniques and acceptable repeatability (less than 3 mm) (Highgenboten et al., 1989) for assessing ligament damage. Further studies should be undertaken to confirm the efficacy of this methodology for quantification of AP laxity in the clinical setting. References Banks, S.A., Hodge, W.A., 1996. Accurate measurement of threedimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Trans. Biomed. Eng. 43, 638–649. Daniel, D.M., Malcom, L.L., Losse, G., Stone, M.L., Sachs, R., Burks, R., 1985. Instrumented measurement of anterior laxity of the knee. J. Bone. Joint Surg. Am. 67, 720–726. Fleming, B.C., Johnson, R.J., Shapiro, E., Fenwick, J., Howe, J.G., Pope, M.H., 1992. Clinical versus instrumented knee testing on autopsy specimens. Clin. Orthop., 196–207.
312
M.T. Thompson et al. / Clinical Biomechanics 19 (2004) 308–312
Fleming, B.C., Peura, G.D., Abate, J.A., Beynnon, B.D., 2001. Accuracy and repeatability of Roentgen stereophotogrammetric analysis (RSA) for measuring knee laxity in longitudinal studies. J. Biomech. 34, 1355–1359. Fleming, B.C., Brattbakk, B., Peura, G.D., Badger, G.J., Beynnon, B.D., 2002. Measurement of anterior–posterior knee laxity: a comparison of three techniques. J. Orthop. Res. 20, 421– 426. Highgenboten, C.L., Jackson, A., Meske, N.B., 1989. Genucom, KT1000, and Stryker knee laxity measuring device comparisons. Device reproducibility and interdevice comparison in asymptomatic subjects. Am. J. Sports Med. 17, 743–746. Karrholm, J., Selvik, G., Elmqvist, L.G., Hansson, L.I., Jonsson, H., 1988. Three-dimensional instability of the anterior cruciate deficient knee. J. Bone Joint Surg. Br. 70, 777–783.
Steiner, M.E., Brown, C., Zarins, B., Brownstein, B., Koval, P.S., Stone, P., 1990. Measurement of anterior–posterior displacement of the knee. A comparison of the results with instrumented devices and with clinical examination. J. Bone Joint Surg. Am. 72, 1307– 1315. Stiehl, J.B., Komistek, R.D., Dennis, D.A., Paxson, R.D., Hoff, W.A., 1995. Fluoroscopic analysis of kinematics after posterior-cruciateretaining knee arthroplasty. J. Bone Joint Surg. Br. 77, 884–889. Uh, B.S., Beynnon, B.D., Churchill, D.L., Haugh, L.D., Risberg, M.A., Fleming, B.C., Un, B.S., 2001. A new device to measure knee laxity during weightbearing and non-weightbearing conditions. J. Orthop. Res. 19, 1185–1191. Wroble, R.R., VanGinkel, L.A., Grood, E.S., Noyes, F.R., Shaffer, B.L., 1990. Repeatability of the KT-1000 arthrometer in a normal population. Am. J. Sports Med. 18, 396–399.
Brief report: validation of a system for automated measurement of knee laxity M.T. Thompson a
a,*
, M.A. Conditt a, S.K. Ismaily a, A. Agarwal b, P.C. Noble
a,b
Institute of Orthopedic Research and Education, 6550 Fannin Street, Suite 2512, Houston, TX 77030, USA b Baylor College of Medicine, Houston, TX 77030, USA Received 24 September 2003; accepted 4 December 2003
Abstract Objective. To determine the accuracy and repeatability of an automated quantitative fluoroscopic imaging system for measuring knee laxity. Design. Cadaveric validation study. Background. Current methods of measuring anterior–posterior laxity lack sufficient accuracy and repeatability. A commercially developed fluoroscopic software package, capable of measuring laxity, required validation. Methods. Five human cadaveric knees were used. A constant force of 130 N was applied anteriorly and posteriorly in turn to the tibia of each knee with the femur fixed in 30 and 90 of flexion. Quantitative fluoroscopic measurements of anterior–posterior laxity were determined using image analysis software. Fluoroscopic results were compared to the true anterior–posterior displacements of the tibia, which were simultaneously recorded using linear transducers directly attached to the cadaveric specimens. Results. The quantitative fluoroscopic method underestimated laxity by an average of 0.40 mm with a root mean square error of 0.49 mm. The 95% confidence intervals for anterior and posterior laxity error were calculated to be )0.99 to 0.25 mm and )0.89 to 0.03 mm, respectively, where a negative error represents an underestimation. Conclusions. The quantitative fluoroscopic method offers a dramatic improvement in accuracy over current laxity measurement techniques and acceptable repeatability for assessing ligament damage. Relevance The considerably more accurate, validated measurement system of this study could improve ligament assessment and diagnosis, and the recognition of injuries otherwise undetected with current methods. 2004 Elsevier Ltd. All rights reserved. Keywords: Knee; Laxity; Fluoroscope; AP drawer
1. Introduction Anterior–posterior (AP) knee laxity is routinely assessed in evaluating knee ligament injuries and the success of treatment. Several mechanical, non-invasive methods have been developed to measure AP laxity using devices, such as arthrometers, attached to the lower leg. Although these devices are commonly used to obtain objective measurements of knee laxity, they have been shown to significantly overestimate true laxity and often
*
Corresponding author. E-mail address: [email protected] (M.T. Thompson).
0268-0033/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.clinbiomech.2003.12.004
exhibit poor repeatability (Uh et al., 2001; Fleming et al., 2002; Fleming et al., 1992; Daniel et al., 1985). The KT-1000 Knee Arthrometer (MedMetrics, San Diego, CA, USA) is the most widely used knee laxity measurement system (Daniel et al., 1985). When externally strapped to the lower leg, the KT-1000 records tibial translations relative to the patella during AP loading. A confidence interval of 2.4 mm within subjects and root mean square (RMS) errors as high as 3.9 mm have been calculated for the KT-1000 when compared to highly accurate (0.2 mm RMS error) Roentgen stereophotogrammetry analysis (RSA) (Fleming et al., 2002; Karrholm et al., 1988; Steiner et al., 1990; Wroble et al., 1990). RSA, however, is an invasive technique, requiring the implantation of tantalum beads into the femur and
M.T. Thompson et al. / Clinical Biomechanics 19 (2004) 308–312
tibia to provide radiographic landmarks imaged with biplanar radiography. Planar stress radiography is another common method to measure laxity using sequential lateral radiographs taken during anterior and posterior shear loads. The relative changes in displacement of the posterior aspect of the tibia are measured relative to the femoral condyles. The inaccuracy of planar stress radiography has been reported as high as 5.0 mm when compared to RSA (Fleming et al., 2002). Computer-assisted, fluoroscopic measurements offer non-invasive, real-time, objective quantification of AP laxity, but are seldom used clinically due to the lack of automated measurement software, poorly developed clinical guidelines that use knee laxity measurements, and concerns relating to radiation exposure. Moreover, though these techniques have been utilized in laboratory studies, they presently lack experimental validation for measurement of AP laxity. Studies using fluoroscopy to measure the complex 3D kinematics of total knee replacements report translational errors of 0.2–0.5 mm and out-of-plane errors as great as 6.0 mm (Stiehl et al., 1995; Banks and Hodge, 1996). Those studies, however, used accurately known 3D geometric properties of the implants or required computerized tomography (CT) of the bones to track in-plane and out-of-plane behavior, rather than tracking the in-plane motion of the femur and tibia directly from the fluoroscopic image. An accurate and reproducible method for measuring knee laxity that can be used in routine clinical practice would facilitate improved treatment guidelines for patients with knee ligament injuries. The accuracy of fluoroscopic measurements can be influenced by several factors including the extent of out-of-plane motion during the laxity exam, and the resolution and quality of the fluoroscope and its calibration equipment. The objective of this study was to determine the accuracy of two-dimensional, sagittal plane knee laxity measurements from a quantitative fluoroscopic imaging system through direct comparison with physical measurements in cadaveric knees.
309
2. Methods Five fresh frozen left cadaveric lower limbs (all males, ages 52–59) with no evidence of ligamentous pathology were cut approximately 30 cm above and below the joint line. The femur and tibia of each specimen was potted in PVC pipe and mounted in a two degree-of-freedom apparatus that allowed only anteroposterior motion of the tibia and superior/inferior motion of the femur (Fig. 1). The device was constructed completely of non-ferrous metal to minimize any distortion of the fluoroscopic image. The continuous AP motions of the femur and tibia were directly measured using two linear displacement transducers (UniMeasure, Inc., Corvallis, OR, USA), aligned in the AP direction and attached to the bones using radiolucent screws. No instrumentation was visible in the fluoroscopic image and no markers or beads were implanted in the bones (Fig. 1). The fluoroscopic imaging system (kimax 1024, Medical Metrics, Inc, Houston, TX, USA, camera resolution: 1024 · 1024 pixels, intensifier resolution: 0.302 mm/pixel) was positioned prior to testing so that the knee joint was centered on the fluoroscopic image. Alternating anterior and posterior forces of 132 N were applied to the tibia in 4-s cycles via a pneumatic cylinder. The knees were allowed to rest unloaded between runs. During each run, anteroposterior displacement of the femur and tibia were recorded at 30 Hz by the linear transducers placed parallel to the image intensifier and simultaneously by the image analysis software from Medical Metrics, Inc. Data collection was synchronized so that for each fluoroscopic frame there was a corresponding transducer value. The knee was loaded for one cycle initially to precondition the knee and obtain a repeatable starting and return point. Three runs with 3–5 cycles per run were performed with each knee held in 30 and 90 of flexion, for both anteriorly and posteriorly directed forces. One recorded fluoroscopic frame was chosen from each cycle of each run for comparison with the transducer measurements, generating a total of 275 laxity measurements. Prior to analysis of the fluoroscopic images, geometric distortion
Fig. 1. The experimental setup and corresponding fluoroscopic image.
310
M.T. Thompson et al. / Clinical Biomechanics 19 (2004) 308–312
transducer-based measurements of laxity of each knee, measured as one standard deviation about the mean, ranged from 0.06 to 1.78 mm for the five specimens. Of the two flexion angles and the two loading directions, the largest average error occurred during posterior loading at 30 (underestimation of 0.62 mm). The largest underestimation and overestimation of laxity were 0.97 and 0.53 mm, respectively. The 95% confidence intervals for anterior and posterior laxity error were calculated to be )0.99 to 0.25 mm and )0.89 to 0.03 mm, respectively, where a negative error represents an underestimation.
due to the image intensifier and the magnetic field of the environment was minimized using a distortion correction algorithm. The AP laxity of the knee was defined as the change in relative position of the tibia with respect to the femur in the anatomic AP direction, before and after loading. From the transducers, laxity was simply calculated by subtracting the femoral displacement from the tibial displacement. From the fluoroscopic images the outline of each bone was traced and anatomic landmarks in the first frame of the fluoroscopic imaging sequence were identified. The outer borders of the tibia and femur were traced based on a generic template provided by the software. The software tracked each bone through all frames of the fluoroscopic imaging sequence based on the unique radiographic appearance of the bone. New coordinates of the anatomic landmarks were calculated for each frame. The landmarks were chosen so that the AP translation of the tibia could be calculated relative to the femur. Any sensitivity of the tracking process to operator variation in tracing the outline of the bones would be minimized by the stabilization feature of the software which holds either the tibia or femur in a fixed position on the screen as the fluoroscopic imaging sequence is viewed. The accuracy of tracking was verified visually using this feature, and minor mistakes in the tracking process could be corrected by the user. The error in each automated laxity measurement was calculated by subtracting the fluoroscopic estimate of AP translation of the tibia with respect to the femur from the true value of AP laxity as recorded by the linear transducers. Paired and unpaired comparisons were performed between laxity, error, loading direction, and flexion angle. Correlation analyses were also performed between fluoroscopic and true measurements, and between laxity and measured error.
4. Discussion In this study, the accuracy of a novel method of measuring the AP laxity of the knee using a non-invasive quantitative fluoroscopic technique was evaluated. The nominal resolution of this system is approximately 0.3 mm per pixel. Larger errors might be expected with a lower resolution imaging system, although the software used to track motion in the fluoroscopic images interpolates between frames, and is thereby capable of giving subpixel accuracy. Both the transducers and the fluoroscopic method used in this study detected significant differences and similar trends in AP laxity between anterior and posterior loads and between 30 and 90 of flexion. On average, the fluoroscopic method underestimated the true laxity by 0.40 (SD, 0.28) mm with an RMS error of 0.49 mm. In laxity measurement, the 95% confidence interval for error is perhaps more relevant. The results have shown that regardless of loading direction, the magnitude of the 95% confidence interval for laxity error is less than 1.26 mm. There was little correlation ()0.314) between AP laxity and AP error, suggesting that fluoroscopic error does not increase with increased laxity. The source of the apparently systematic error was not clear, since both the fluoroscopic imaging system and the linear transducers were calibrated. The linear displacement transducers used to measure the true AP displacements have essentially infinite resolution with a repeatability of 0.03% full scale (0.02 mm) and a linearity of 1% full scale (0.50 mm). The data acquisition system limited the obtainable transducer resolution to approximately 0.01 mm. One possible source of the
3. Results The fluoroscopic measurement system slightly underestimated AP laxity for both anterior and posterior loading, and at both flexion angles of 30 and 90 (Table 1). The average error across all specimens and both loading directions was calculated to be an underestimation of 0.40 (SD, 0.28) mm. Variability in the
Table 1 Average values of AP laxity using direct transducer measurement and using the fluoroscopic system during anterior and posterior loading, and at 30 and 90 of flexion True laxity Fluoro laxity Error P value
Anterior
SD
Posterior
SD
P value
30
SD
90
SD
P value
3.00 2.62 )0.38 <0.01
1.23 1.23
3.33 2.90 )0.43 <0.01
1.36 1.27
0.03 0.05
3.46 2.92 )0.54
2.16 1.89
2.89 2.61 )0.28
1.03 1.12
<0.01 0.04
All values are in mm.
M.T. Thompson et al. / Clinical Biomechanics 19 (2004) 308–312
systematic error was the miscalculation of the magnification in the fluoroscopic images, which was estimated based on the distance of the midsagittal plane of the knee from the plane of the image intensifier, measured using a standard metric ruler. Clearly, if the systematic errors were removed, the error in the fluoroscopic measurements would be significantly less. A systematic, empirical-based correction could be applied to the estimated magnification to reduce subjective errors in the measurements. Since the soft tissue of the human knee can respond differently to even consecutive loads of the same magnitude, repeatability of AP laxity measurement using the fluoroscopic method could not be uniquely determined in this study. Perhaps more representative of quantitative fluoroscopic repeatability is one standard deviation of error in the laxity measurement (0.28 mm). This suggests that although the fluoroscopic method does underestimate laxity, its accuracy varies only slightly between measurements. There was no significant difference in the error of laxity measurements recorded in response to anterior and posterior loading, however, a significant difference (P < 0:001) was found between the 30 and 90 flexion angles ()0.54 vs. )0.27 mm, respectively). Previous work has assessed the accuracy of mechanical methods of measuring the AP laxity of the knee such as the KT-1000 arthrometer and the Vermont knee laxity device (VKLD) (Uh et al., 2001; Fleming et al., 2002; Steiner et al., 1990; Wroble et al., 1990). Fleming et al. (2002) performed AP laxity measurements on 15 subjects and compared the results obtained using the KT-1000, planar stress radiography, and RSA. Assuming RSA to be the true value, RMS errors of 5.0 and 3.9 mm were calculated for the KT-1000 and planar stress radiography, respectively. In another study, the repeatability of laxity measurements recorded by experienced operators applying a 130 N load was found to be 1.4 (SD, 1.4 mm) with the KT-1000 and as 2.4 (SD, 2.5 mm) with the VKLD (Uh et al., 2001). Quantitative fluoroscopic errors determined in previous studies examining total knee replacement kinematics (Stiehl et al., 1995; Banks and Hodge, 1996) are not relevant in fluoroscopic measurement of AP laxity. This study has several limitations. Because of the alignment and constraints of the testing device, the tibia was only free to move in the AP direction, thus, measurement of out of plane motion was unnecessary and the linear transducers were sufficient. In a clinical setting, however, out of plane motion may be unavoidable, and the errors due to out-of-plane motion remain to be characterized. The testing device also applied repeatable, controlled loads to the tibia. To achieve a clinical test for knee laxity that minimizes variability due to the individual performing the test, a mechanism to apply controlled loads to the tibia would be of benefit. Such a device is not readily available for clinical use. There are
311
also knee ligament, meniscal and cartilage injuries that may not be detectable with a two-dimensional test performed under fluoroscopic imaging. Finally, there was minimal rotation of the femur or tibia during the cadaver loading experiments. In clinical practice, sagittal plane rotations might occur, and knee laxity is poorly defined in the presence of combined AP translation and sagittal plane rotation. The clinical value of more accurate knee laxity measurements is not well-defined. Is it really advantageous to have knee laxity measurements that are accurate to less than 1 mm compared to measurements that are accurate to 3 or more mm? Perhaps with improved accuracy, it may be possible to detect a partial ligament injury, or detect a ligament injury when swelling or muscle action is limiting motion, or detect changes in laxity following surgery that may predict if the reconstruction is failing. The improved accuracy would at least facilitate laboratory and clinical experiments to address the above hypotheses. Perhaps more importantly, knee laxity measurements that are minimally effected by inter-tester variation and are free from errors due to soft-tissue artifacts could be used by general practitioners to determine which patients need to be seen by an orthopedic surgeon.
5. Conclusion The quantitative fluoroscopic technique evaluated in this study underestimated knee laxity by an average of 0.40 (SD, 0.28) mm for combined anterior and posterior directions, with a 95% confidence level of 1.09 mm and an RMS error of 0.49 mm. In comparison, the RMS errors of mechanical arthrometers utilized clinically is reported to be as high as 3.9 mm (Uh et al., 2001; Fleming et al., 2002; Daniel et al., 1985; Fleming et al., 2001). In view of these findings, the quantitative fluoroscopic method offers a dramatic improvement in accuracy over current AP laxity techniques and acceptable repeatability (less than 3 mm) (Highgenboten et al., 1989) for assessing ligament damage. Further studies should be undertaken to confirm the efficacy of this methodology for quantification of AP laxity in the clinical setting. References Banks, S.A., Hodge, W.A., 1996. Accurate measurement of threedimensional knee replacement kinematics using single-plane fluoroscopy. IEEE Trans. Biomed. Eng. 43, 638–649. Daniel, D.M., Malcom, L.L., Losse, G., Stone, M.L., Sachs, R., Burks, R., 1985. Instrumented measurement of anterior laxity of the knee. J. Bone. Joint Surg. Am. 67, 720–726. Fleming, B.C., Johnson, R.J., Shapiro, E., Fenwick, J., Howe, J.G., Pope, M.H., 1992. Clinical versus instrumented knee testing on autopsy specimens. Clin. Orthop., 196–207.
312
M.T. Thompson et al. / Clinical Biomechanics 19 (2004) 308–312
Fleming, B.C., Peura, G.D., Abate, J.A., Beynnon, B.D., 2001. Accuracy and repeatability of Roentgen stereophotogrammetric analysis (RSA) for measuring knee laxity in longitudinal studies. J. Biomech. 34, 1355–1359. Fleming, B.C., Brattbakk, B., Peura, G.D., Badger, G.J., Beynnon, B.D., 2002. Measurement of anterior–posterior knee laxity: a comparison of three techniques. J. Orthop. Res. 20, 421– 426. Highgenboten, C.L., Jackson, A., Meske, N.B., 1989. Genucom, KT1000, and Stryker knee laxity measuring device comparisons. Device reproducibility and interdevice comparison in asymptomatic subjects. Am. J. Sports Med. 17, 743–746. Karrholm, J., Selvik, G., Elmqvist, L.G., Hansson, L.I., Jonsson, H., 1988. Three-dimensional instability of the anterior cruciate deficient knee. J. Bone Joint Surg. Br. 70, 777–783.
Steiner, M.E., Brown, C., Zarins, B., Brownstein, B., Koval, P.S., Stone, P., 1990. Measurement of anterior–posterior displacement of the knee. A comparison of the results with instrumented devices and with clinical examination. J. Bone Joint Surg. Am. 72, 1307– 1315. Stiehl, J.B., Komistek, R.D., Dennis, D.A., Paxson, R.D., Hoff, W.A., 1995. Fluoroscopic analysis of kinematics after posterior-cruciateretaining knee arthroplasty. J. Bone Joint Surg. Br. 77, 884–889. Uh, B.S., Beynnon, B.D., Churchill, D.L., Haugh, L.D., Risberg, M.A., Fleming, B.C., Un, B.S., 2001. A new device to measure knee laxity during weightbearing and non-weightbearing conditions. J. Orthop. Res. 19, 1185–1191. Wroble, R.R., VanGinkel, L.A., Grood, E.S., Noyes, F.R., Shaffer, B.L., 1990. Repeatability of the KT-1000 arthrometer in a normal population. Am. J. Sports Med. 18, 396–399.