- Email: [email protected]
Renal Transplant Elasticity Ultrasound Imaging: Correlation Between Normalized Strain and Renal Cortical Fibrosis
Ultrasound in Med. & Biol., Vol. 39, No. 9, pp. 1536–1542, 2013 Copyright Ó 2013 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/$ - see front matter
http://dx.doi.org/10.1016/j.ultrasmedbio.2013.04.007
Original Contribution
d
RENAL TRANSPLANT ELASTICITY ULTRASOUND IMAGING: CORRELATION BETWEEN NORMALIZED STRAIN AND RENAL CORTICAL FIBROSIS JING GAO,* WILLIAM WEITZEL,y JONATHAN M. RUBIN,z JAMES HAMILTON,x JUN LEE,{ DARSHANA DADHANIA,{ and ROBERT MIN* y
* Department of Radiology, New York-Presbyterian Hospital, Weill Cornell Medical College, New York, New York, USA; Department of International Medicine, University of Michigan Hospital and VA Medical Center, Ann Arbor, Michigan, USA; z Department of Radiology, University of Michigan Hospital, Ann Arbor, Michigan, USA; x Epsilon Imaging, Ann Arbor, Michigan, USA; and { Rogosin Institute, New York-Presbyterian Hospital, Weill Cornell Medical College, New York, New York, USA (Received 20 November 2012; revised 18 March 2013; in final form 11 April 2013)
Abstract—After transplantation, over a widely variable time course, the cortex of the transplanted kidney becomes stiffer as interstitial fibrosis develops and renal function declines. Elasticity ultrasound imaging (EUI) has been used to assess biomechanical properties of tissue that change in hardness as a result of pathologic damage. We prospectively assessed the hardness of the renal cortex in renal transplant allograft patients using a normalized ultrasound strain procedure measuring quasi-static deformation, which was correlated with the grade of renal cortical fibrosis. To determine cortical strain, we used 2-D speckle-tracking software (EchoInsight, Epsilon Imaging, Ann Arbor, MI, USA) to perform offline analysis of stored ultrasound loops capturing deformation of renal cortex and its adjacent soft tissue produced by pressure applied using the scanning transducer. Normalized strain is defined as the mean developed strain in the renal cortex divided by the overall mean strain measured in the soft tissues from the abdominal wall to pelvic muscles. Using the Banff scoring criteria for renal cortical fibrosis as the gold standard, we classified 20 renal transplant allograft biopsy tissue samples into two groups: group 1 (n 5 10) with mild (,25%) renal cortical fibrosis and group 2 (n 5 10) with moderate (26%–50%) renal cortical fibrosis. An unpaired two-tailed t-test was used to determine the statistical difference in strains between patients with mild and those with moderate renal cortical fibrosis. Receiver operating characteristic curve analysis was performed to assess the accuracy of developed strain and normalized strain in predicting moderate renal cortical fibrosis. The reference strain did not significantly differ between the two groups (p 5 0.10). However, the developed renal cortical strain in group 1 with mild fibrosis was higher than that in group 2 with moderate fibrosis (p 5 0.025). The normalized strain in group 1 was also higher than that in group 2 (p 5 0.0014). The areas under receiver operating characteristic curves for developed strain and normalized strain were 0.78 and 0.95, respectively. The optimal cutoff for distinguishing moderate renal cortical fibrosis was 20.08 for developed strain (sensitivity 5 0.50, specificity 5 1.0) and 2.5 for normalized strain (sensitivity 5 0.80, specificity 5 1.0). In summary, renal cortex strain is strongly correlated with grade of renal cortical fibrosis. Normalized strain is superior to developed strain in distinguishing moderate from mild renal cortical fibrosis. We conclude that free-hand real-time strain EUI may be useful in assessing the progression of cortical fibrosis in renal transplant allografts. Further prospective study using this method is warranted. (E-mail: [email protected]) Ó 2013 World Federation for Ultrasound in Medicine & Biology. Key Words: Elastography, Renal cortical fibrosis, Renal transplant, Ultrasound strain.
ratory tests and, if definitive diagnosis is needed, renal biopsy. However, routine laboratory tests such as serum creatinine are insensitive to loss of renal function (Hunsicker and Bennett 1995), and kidney biopsy is limited by its invasive nature and high cost (Schwarz et al. 2005). Therefore, improved non-invasive techniques to assess renal function and detect renal dysfunction early would be desirable. Elasticity ultrasound imaging (EUI) is an emerging non-invasive imaging
INTRODUCTION In current practice, assessment of renal transplant dysfunction relies on the monitoring of biochemical labo-
Address correspondence to: Jing Gao, Department of Radiology, New York-Presbyterian Hospital, Weill Cornell Medical College, 525 East 68th Street, Suite 8A, New York, NY 10065, USA. E-mail: [email protected] 1536
Renal transplant ultrasound elastography d J. GAO et al.
technique for the assessment of internal organ and tissue biomechanical properties that may be useful in assessing pathologic changes that influence tissue mechanics (O’Donnell et al. 1993; Xu et al. 2012). In the kidney, EUI may provide information about the mechanical changes that accompany histologic changes useful for monitoring renal disease (Chaturvedi et al. 1998). Quasi-static ultrasound strain measurements of the renal cortex in renal transplant recipients may help detect and monitor tissue hardness changes in a chronic transplant allograft nephropathy (CTAN) (Weitzel et al. 2004). More recently, techniques such as acoustic radiation force impulse imaging (Stock et al. 2011; Syversveen et al. 2011, 2012), shear wave velocity imaging (Gennisson et al. 2010; Ozkan et al. 2013) and shear wave dispersion vibrometry (Amador et al. 2009) have been used to assess renal cortical mechanical changes in renal transplants. Work remains to establish and improve the reliability and accuracy of these techniques, taking into account the effects of renal anisotropy and boundary conditions (Gennisson et al. 2012; Rubin et al. 1988; Syversveen et al. 2011). We chose to test and further evaluate the free-hand push method suggested in earlier work (Weitzel et al. 2004) because of (i) the locally isotropic renal cortical structure, wherein pathologic changes of interest develop relatively near the surface (transducer); (ii) the possibility of standardized data collection in the measurement by normalizing with a reference strain measurement from the acquired strain image; (iii) the availability of gold standard criteria for reference measurements using an accepted pathology scoring system for fibrosis in the transplant setting; and (iv) the simplicity of the measurement method, which would allow for broad clinical use if the method should prove useful. In this study we sought to prospectively assess, using EUI, the relationship between the hardness of renal cortex, represented by mean developed renal cortical strain measurements and normalized strain measurements (developed renal cortical strain divided by imaged soft tissue strain), and the grade of renal cortical fibrosis by Banff criteria. The ultimate goal of this study was to assess the value of ultrasound strain imaging in monitoring the progression of renal cortical fibrosis in the allograft after transplantation. METHODS Patients We performed EUI on 20 patients (11 men and 9 women, age range 28–87 y, mean age 50 6 16 y) who underwent transplanted kidney biopsy from March 2012 to August 2012. All patients were studied after providing written informed consent in this study. All color Doppler
1537
sonography and quasi-static elastography were conducted at the New York-Presbyterian Hospital, Weill Cornell Medical College, after obtaining approval from institutional review board of Weill Cornell Medical College. The study was compliant with the Health Insurance Portability and Accountability Act. Patients were eligible to participate if a referring nephrologist or transplant surgeon referred them for renal biopsy. The indications for renal biopsy were loss of renal function, suspicion of allograft rejection, or participation in protocol renal transplant biopsy. As a standard of care for patients having renal transplant in our institution, color Doppler sonography was routinely performed to screen for allograft vascular complications, for example, transplant renal artery stenosis, and non-vascular complications, such as hydronephrosis for an elevated serum creatinine. Patients with hydronephrosis, large perinephric collections, immediate post-transplantation status (,7 d after the surgery), significant transplant renal artery stenosis and an existing intrarenal vascular abnormality (e.g., arteriovenous fistula), which may be contraindications for renal biopsy and/or affect renal cortical strain measurement, were not enrolled in the study; in addition, patients who could not tolerate compression for any reason were excluded. Real-time ultrasound data acquisition Two investigators who had more than 20 y of experience in ultrasound scanning and were trained in freehand compression performed the scans (one investigator scanned 15 cases, and the other scanned 5 cases). Gentle but firm free-hand compression of a transplanted kidney was performed during standard renal transplant sonography before kidney biopsy, on the same day as the biopsy (n 5 17) or ,7 d before the biopsy (n 5 3). Subjects were placed in the supine position. The transplanted kidney was imaged using either a General Electric (Logic E9, General Electric, Milwaukee, WI, USA) or Siemens (Sequoia 512, Siemens Medical Solution, Mountain View, CA, USA) scanner equipped with a multi-frequency 2- to 4-MHz curved linear array transducer. Transmission gel was placed on the anterior abdominal wall as standard acoustic coupling for ultrasound examination over the region where the transplanted kidney is located. The speckle reduction setting on the Logic E9 scanner was turned off while capturing realtime compression to increase the fidelity of speckle data acquisition. The speckle reduction feature performs an averaging function on the images, making the examination more pleasing to the eye, but may influence tracking and therefore was disabled as part of this protocol. We used a transducer frequency range of 2–4 MHz to improve penetration of the ultrasound beam to the deep pelvis. The depth of gray-scale imaging ranged
1538
Ultrasound in Medicine and Biology
Volume 39, Number 9, 2013
Fig. 1. (a) Real-time gray-scale image of free-hand compression on the short axis of a transplanted kidney recorded prebiopsy. The image was from a case in group 1 (,25% renal cortical fibrosis), a patient who had undergone transplantation of a living non-related donor kidney 4 mo earlier. Regional strain measurements were performed using connected dual regions of interest (ROIs), which are axially placed with respect to the center of the transducer (cyan and red dotted lines). The movement of each ROI of the tethered pair through the image loop is calculated from tissue motion estimates produced by speckle tracking. Developed strain (the line between the two cyan arrows) is the strain in renal cortex, and reference strain (the line between the two red arrows) is the strain in the soft tissue from the anterior abdominal wall to pelvic muscles. (b) Graph of the strain measurements from (a). Developed strain (cyan solid line) is significantly higher than reference strain (red solid line). Normalized strain 5 developed strain/reference strain. In this case, developed strain, the reference strain and normalized strain are 20.10, 20.02 and 5, respectively. Histopathologic analysis of the kidney biopsy specimen revealed ,5% interstitial fibrosis, activated lymphocytes, marked active tubulitis and interstitial inflammation.
from 9 to 13 cm, as a transplanted kidney is usually located superficially in the right or left lower quadrant just under the anterior abdominal wall. We used conventional gray-scale imaging to select the region from the skin of the anterior abdominal wall to pelvic muscles that was free of bowel and urinary bladder after routinely imaging a transplanted kidney. The conventional ultrasound examination included gray-scale, color Doppler and spectral Doppler analysis. We started an applied axial compression (Kallel et al. 1997) by pressing on the anterior abdominal wall and underlying tissue including the transplanted kidney and pelvic muscles. The pushing force applied was similar to deep palpation in a physical examination, but the ultrasound transducer was used while monitoring the subject to ensure his or her comfort during the ultrasound palpation examination. The pushing force was directed toward the iliac, and the deformation was measured down to the iliopsoas muscle in the pelvis. A compression was considered effective when the deformation in the region of interest reached 10% of the initial depth of soft tissue (Emelianov et al. 1995; Xu et al. 2012) measured from the abdominal wall to the iliopsoas muscle. Compression was conducted in sagittal and/or transverse views of the transplanted kidney. On conclusion of the application of constant pressure on the kidney, the pressure was quickly released while ultrasound data capture continued to provide pre- and
post-push reference images for post-procedure data processing and speckle tracking (Emelianov et al. 1998). To improve the accuracy and robustness of strain measurements, the acquisition frame rate was increased to reduce speckle de-correlation between frames. Through use of a single transmit focus and reduced spatial resolution (i.e., beam density), frame rates of 33–47 Hz were achieved for full field-of-view imaging covering the longitudinal or transverse section of the kidney. In addition, the ultrasound probe was carefully controlled to apply compression normal to the transducer face, further limiting speckle de-correlation by minimizing out-of-plane tissue motion. All captured images were stored in the Picture Achieving and Communications System (PACS) in the Department of Radiology. 2-D speckle tracking for measuring renal transplant strains Gray-scale image loops were captured and stored in Digital Imaging and Communications in Medicine (DICOM) format and pre-processed by speckle tracking software (EchoInsight, Epsilon Imaging, Ann Arbor, MI, USA) to prepare for strain analysis. For this study, regional strain measurements were done using a connected, dual region of interest (ROI), as shown by the cyan- and red-colored graphics in Figures 1(a, b) and 2
Renal transplant ultrasound elastography d J. GAO et al.
1539
Fig. 2. (a) Image of free-hand compression on a longitudinal section of a transplanted kidney from a case in group 2 (26%–50% renal cortical fibrosis), who had an elevated creatinine level (3.2 mg/dL) in his fourth year after renal transplantation. Strain measurements were performed along axial of the cortex at the center of the transducer (cyan dotted line and red dotted line). Care should be taken to avoid calyces and renal pelvis (red arrow) and medulla pyramid (white arrow), which may introduce error into measurements of developed strain. (b) Graph of the strain measurements from (a). In this case, developed strain (cyan line), reference strain (red line) and normalized strain (developed strain/reference strain) are 20.10, 20.08 and 1.25, respectively. Histopatholic analysis of the kidney biopsy specimen revealed 30% interstitial fibrosis, active moderate tubular atrophy and chronic active antibodymediated rejection. It is clear from this graph that the difference between developed strain and reference strain is not significantly different in this case of moderate renal cortical fibrosis. Compared with the case of mild renal cortical fibrosis in Figure 1, although using a relatively higher reference strain (20.08 vs. 20.02), normalized strain in this case of moderate renal cortical fibrosis is significantly lower than that in mild renal cortical fibrosis (1.25 vs. 5).
(a, b). The movement of each ROI of the tethered pair through the image loop is calculated from tissue motion estimates produced by speckle tracking. The relative motion between the ROIs is expressed as strain, and represents the average tissue deformation between the ROI pair (i.e., total strain along the connecting line of the ROI tether). In this study, developed strain is defined as the strain in the renal cortex between the renal capsule and the collecting system (range, 0.8–1.5 cm in anteriorposterior dimension, shown as a cyan line in Fig. 1a, b), and reference strain is total deformation from the skin of the abdominal wall to the pelvic muscles after compression (shown as a red line in Fig. 1a, b). Care should be taken to avoid capsule, calyx, medulla and sinus fat that may cause an error in measuring developed strain (Fig. 2a). Preferably, strain is measured along the direction of compression, in this case the normal from the transducer surface at the center of the image (i.e., vertical direction in Fig. 2a). The developed strain (cyan) and reference region (red) during compression are indicated in Figures 1b and 2b. All strain measurements for this study were made by a single operator (J.G.). For each subject, both developed strain and reference strain were measured using these methods.
Kidney biopsy and histopathology Ultrasound-guided transplanted kidney biopsy was performed by the nephrologist using our standard protocol. Patients fasted 8–10 h before the kidney biopsy. The procedure was performed with patients in the supine position. A Siemens scanner equipped with a 4V1 sector array transducer (Sequoia 512, Siemens Medical Solution) with biopsy guide (Ultra-Pro II Needle Guide, CIVCO Medical Instruments, Kalona, IA, USA) in a sterile transducer cover provided real-time, gray-scale images to guide local anesthesia and kidney biopsy. The mid- to lower pole of the kidney was commonly selected for kidney biopsy, and an area lacking prominent vessels on sonography was chosen as the biopsy site. To obtain an adequate biopsy specimen and avoid damaging renal vessels, the angle between the biopsy needle and the skin was 18 or 32 depending on the thickness of the renal cortex and existing vessels along the biopsy path. An 18-gauge biopsy needle (Bard Peripheral Technologies, Covington, GA, USA) was inserted and advanced to the kidney cortex under ultrasound guidance after local anesthesia. Static images and cine loops of renal sonography and biopsy were stored in the PACS system in the Department of Radiology.
1540
Ultrasound in Medicine and Biology
Volume 39, Number 9, 2013
Table 1. Clinical information and ultrasound strain on 20 renal transplants
Male/female ratio Live donor/deceased donor Age Duration of transplant (mo) Creatinine (ng/dL) Glomerular filtration rate Reference strain Developed strain Normalized strain*
Group 1 (,25% cortical fibrosis)
Group 2 (36%–50% cortical fibrosis)
p (t-test)
3/7 7/3 46 6 16 21.3 6 27.7 1.89 6 1.16 42.81 6 6.8 20.03 6 0.02 20.14 6 0.05 5.23 6 2.64
8/2 1/9 58 6 12 46.2 6 37.1 3.49 6 0.81 18.8 6 6.48 20.05 6 0.09 20.08 6 0.04 1.73 6 0.83
0.07 0.065 0.002 0.0005 0.100 0.025 0.0014
* Normalized strain 5 developed strain/reference strain.
A pathologist with more than 25 y of experience in the interpretation of transplanted kidney biopsies reviewed biopsy specimens; she was blinded to the results of renal sonography. Biopsies were evaluated on the basis of the Banff’07 classification of renal allograft pathology (Solez et al. 2008). Further, the glomeruli, tubules, interstitium and vessels were evaluated. Special stains used were periodic acid-Schiff, periodic acid-silver methenamine and trichrome. The severity of interstitial fibrosis/tubular atrophy was estimated to the nearest 5% and then divided into four categories: 0, no fibrosis present; 1, mild (,25%) fibrosis; 2, moderate (26%–50%) fibrosis; and 3, severe (.50%) fibrosis.
The difference in reference strain between the two groups did not reach statistical significance (p 5 0.10), and of note, this reference strain was lower in the lowfibrosis group than in the high-fibrosis group. The differences in developed strain and normalized strain between the two groups were statistically significant. Developed strain in group 1 with mild renal cortical fibrosis was higher than that in group 2 with moderate renal cortical fibrosis (p 5 0.025). Developed cortical strain was greater in group 1 even though the overall (background) strain was lower in this group compared with group 2. Further, normalized strain was significantly higher in group 1 than in group 2 (p 5 0.0014). The area under the ROC curve for developed strain was 0.78 (Fig. 3), whereas that for normalized strain was 0.95 (Fig. 4).
Statistical analysis All variables including developed strain (i.e., strain in renal cortex), reference strain and normalized strain are expressed as means and standard deviations (SD). An unpaired two-tailed t-test (Microsoft Excel) was used to analyze the difference in strain between group 1 with ,25% renal cortical fibrosis and group 2 with 26%–50% renal cortical fibrosis based on Banff cortical fibrosis score criteria. A p value ,0.05 was considered statistically significant. Diagnostic sensitivity and specificity were evaluated on the basis of the receiver operating characteristic curve (ROC) analysis (www.rad.jhmi.edu/ jeng/javarad/roc). The optimal cutoff values were chosen to maximize the sum of sensitivity and specificity. RESULTS Characteristics, including age, gender, duration of transplant, donor status and renal function tests, of 20 patients are listed in Table 1. There were no significant differences in patient age or duration of transplant between the two groups. There were significant differences in serum creatinine (p , 0.002) and glomerular filtration rate (GFR) (p , 0.0005) between the two groups.
Fig. 3. Receiver operating characteristic (ROC) curve analysis was performed to assess the accuracy of developed strain in predicting moderate renal cortical fibrosis. The area under the curve is 0.78. For developed strain, the optimal cutoff for distinguishing moderate renal cortical fibrosis is 20.08 (sensitivity 5 0.50 and specificity 5 1.0).
Renal transplant ultrasound elastography d J. GAO et al.
Fig. 4. Receiver operating characteristic (ROC) curve analysis was performed to assess the accuracy of normalized strain in predicting moderate renal cortical fibrosis. The area under the curve is 0.95. For normalized strain, the optimal cutoff for distinguishing moderate renal cortical fibrosis is 2.5 (sensitivity 5 0.80 and specificity 5 1.0).
Normalized strain appears to have better diagnostic value than developed strain (0.95 vs. 0.78) in moderate renal cortical fibrosis. The optimal cutoff values for detecting moderate renal cortical fibrosis were 20.08 (sensitivity 50% and specificity 100%) and 2.5 (sensitivity 80% and specificity 100%) for developed strain and normalized strain, respectively (Table 2).
DISCUSSION In quasi-static elastography, echo signals acquired before and after a small applied deformation (about 1% inter-frame compression) are correlated to estimate tissue displacements (Varghese 2009). The resulting local tissue displacements are used to estimate deformation by calcuTable 2. Receiver operating characteristic curves (one observer)
Area under receiver operating characteristic curve Cutoff value* Sensitivity Specificity Positive predictive value Negative predictive value 95% Confidence interval Significance level, p
Developed strain
Normalized strain
0.78
0.95
20.08 50% 100% 100% 66.7% 51%–91% 0.025
2.5 80% 100% 100% 88.3% 75%–99% 0.0014
* Cutoff value is applied to distinguish moderate from mild renal cortical fibrosis.
1541
lating strain gradients from the tissue motion field. EUI techniques are based on the hypothesis that soft tissues deform more than stiffer tissues, providing an estimate of tissue stiffness. In free-hand EUI, even if the applied surface stress is kept constant, the subsurface stress may vary from patient to patient and from one examination to another in the same patient. When the subsurface stress is uncertain, methods need to be derived to fairly compare the resulting strain. For any EUI technique to be clinically useful, and in the absence of a full knowledge of the conditions needed to determine Young’s modulus, a simple method of gauging the relative stress (force) between examinations is needed. We chose to normalize the developed strain (deformation) in the renal cortex by the background strain as described under Methods. Although not fully accounting for many variables that influence strain, this method of normalizing strain did increase the sensitivity and specificity of the strain measurement in distinguishing moderate from mild renal cortical fibrosis in our population. There are other methods of strain normalization that may be effective (Weitzel et al. 2004). Although current results suggest that normalization is important to improve the diagnostic accuracy of the measurement, the optimal method of normalization remains to be determined. Previously published methods of normalization based on the entire cross section of the imaged kidney (Weitzel et al. 2004) may have advantages in determining the strain relative to the overall strain of the kidney; however, such methods require the entire kidney to be in the ultrasound field of view throughout the compression measurement. The normalized cortical strain in this study has the practical advantage that only the region of interest (transplanted kidney cortex) and region of reference (soft tissue from the abdominal wall to pelvic muscles) must remain in the field of view during compression. It should be noted that may influence measurement beyond the linearity and normalizing assumptions we applied in this study. For example, tissues are inherently non-linear (Emelianov et al. 1995; Xu et al. 2012) in their stress-strain characteristics. For high static strain, there is reduced incremental strain with additional applied stress as compared with low static strain. This non-linear stressstrain behavior may decrease the diagnostic discrimination of the test. Concurrently, increased strain may at the same time lead to patient discomfort. Conversely, if too little pressure is applied, there may not be enough strain developed in either group to make meaningful measurements. We observed mean strain values of 20.109 that both kept patients perfectly comfortable during the examination and still allowed discrimination between degrees of fibrosis. Although these results were achieved with an examination by an experienced
1542
Ultrasound in Medicine and Biology
operator in this study, standardization of the relationship between developed strain to applied stress and standardized strain targets during the examination may allow for even greater combined sensitivity and specificity, while minimizing or accounting for non-linear effects and maintaining a safe and comfortable EUI examination in future studies and, eventually, in clinical practice. There are limitations to this study. (i) The number of cases enrolled in the study was small. (ii) Tests of intraand inter-observer variation need to be investigated, as only one operator performed only one measurement of strain in this study. (iii) Different factors could potentially affect the accuracy of renal strain measurements, including changes in depth of the ROI and geometry of acquisition, that is, curved array versus linear array transducers. Such changes require further assessment. (iv) Pathologic changes in renal allografts are generally complex, in which renal cortical fibrosis is a marker of prognosis after transplantation. Renal cortical fibrosis may have been a major cause of transplant dysfunction at the time of data collection in this study. However, there are other factors that could result in renal transplant dysfunction such as cellular rejection, mediated rejection and nephrotoxicity. The correlations between ultrasound strain and those pathologic factors require further investigation. In summary, the hardness of the renal cortex, measured as developed strain and normalized strain using the EUI method with ultrasound palpation, is strongly correlated with the degree of renal cortical fibrosis. This method or similar methods may prove to be a valuable tool in the evaluation of renal transplant patients. More work is needed using these or other standardized ultrasound strain data acquisition methods to perform prospective clinical studies to determine the optimal role of EUI in managing renal transplant recipients. REFERENCES Amador C, Urban MW, Warner LV, Greenleaf JF. In vitro renal cortex elasticity and viscocity measurements with shearwave dispersion ultrasound vibrometry (DSUV) on swine kidney. Conf Proc IEEE Eng Med Biol Soc 2009;2009:4428–4431. Chaturvedi P, Insana MF, Hall TJ. Ultrasouns and elasticity imaging to model disease-induced changes in soft-tissue structure. Med Image Anal 1998;2:325–338. Emelianov SY, Erkamp RQ, Lubinski MA, Skovoroda AR, O’Donnell M. Non-linear tissue elasticity: Adaptive elasticity imaging for large
Volume 39, Number 9, 2013 deformations. In: Proceedings, 1998 IEEE Ultrasonics Symposium, Sendai, Japan, October 5–8, 1998:1753–1756. Emelianov SY, Lubinski MA, Weitzel WF, Wiggins RC, Skovoroda AR, O’Donnell M. Elasticity imaging for early detection of renal pathology. Ultrasound Med Biol 1995;21:871–883. Gennisson JL, Grenier N, Hubrecht R, Couzy L, Delmas Y, Derieppe M, Lepreaux S, Merville P, Criton A, Bercoff J, Tanter M. Multiwave technology introducing shear wave elastography of the kidney: Pre-clinical study on a kidney fibrosis model and clinical feasibility study on 29 human renal transplants. In: Proceedings, 2010 IEEE Ultrasonics Symposium October 11–14, 2010:1356–1359. Gennisson JL, Grenier N, Combe C, Tanter M. Supersonic shear wave elastography of in vivo pig kidney: Influence of blood pressure, urinary pressure and tissue anisotropy. Ultrasound Med Biol 2012; 38:1559–1567. Hunsicker LG, Bennett LE. Design of trials of methods to reduce late renal allograft loss: The price of success. Kidney Int Suppl 1995; 52:S120–S123. Kallel F, Ophir J, Maguee K, Krouskop T. Quantitative imaging of lowcontrast elastic modulus distribution in tissues using elastography. In: Proceedings, 1997 IEEE Ultrasonics Symposium, Toronto, ON, October 5–8, 1997:1439–1442. O’Donnell M, Emelianov SY, Skovoroda AR, Lubinski MA, Weitzel WF, Wiggins RC. Quantitative elasticity imaging. In: Proceedings, 1993 IEEE Ultrasound Symposium, Baltimore, MD, October 31–November 3, 1993:893–902. Ozkan F, Yavuz YC, Inci MF, Altunoluk B, Ozcan N, Yuksel M, Sayarlioglu H, Dogan E. Interobserver variability of ultrasound elastography in transplant kidneys: Correlations with clinical-Doppler parameters. Ultrasound Med Biol 2013;39:4–9. Rubin JM, Carson PL, Meyer CR. Anisotropic ultrasonic backscatter from the renal cortex. Ultrasound Med Biol 1988;14:507–511. Schwarz A, Gwinner W, Hiss M, Radermacher J, Mengel M, Haller H. Safety and adequacy of renal transplant protocol biopsies. Am J Transplant 2005;5:1992–1996. Solez K, Colvin RB, Racusen LC, Haas M, Sis B, Mengel M, et al. Banff ’07 classification of renal allograft pathology: Updates and future directions. Am J Transplant 2008;8:753–760. Stock EF, Klein BS, Vo Cong MT, Regenbogen C, Kemmner S, Buttner M, Wagenpfei S, Matevossian E, Renders L, Heemann U, Kuchle C. ARFI-based tissue elasticity quantification and kidney graft dysfunction: First clinical experiences. Clin Hemorheol Microcirc 2011;49:527–535. Syversveen T, Brabrand K, Midtvedt K, Strom EH, Hartmann A, Jakobsen JA, Berstad AE. Assessment of renal allograft fibrosis by acoustic radiation force impulse quantification: A pilot study. Transplant Int 2011;24:100–105. Syversveen T, Midtvedt K, Berstad AE, Brabrand K, Strom EH, Abildgaard A. Tissue elasticity estimated by acoustic radiation force impulse quantification depends on the applied transducer force: An experimental study in kidney transplant patients. Eur Radiol 2012; 22:2130–2137. Varghese T. Quasi-static ultrasound elastography. Ultrasound Clin 2009; 4:323–338. Weitzel WF, Kim K, Rubin JM, Wiggins RG, Xie H, Chen X, Emelianov SY, O’Donnell M. Feasibility of applying ultrasound strain imaging to detect renal transplant chronic allograft nephropathy. Kidney Int 2004;65:733–736. Xu J, Tripathy S, Rubin JM, Stidham RW, Johnson LA, Higgins PD, Kim K. A new nonlinear parameter in the developed strain-toapplied strain of the soft tissue and its application in ultrasound elasticity imaging. Ultrasound Med Biol 2012;38:511–523.
http://dx.doi.org/10.1016/j.ultrasmedbio.2013.04.007
Original Contribution
d
RENAL TRANSPLANT ELASTICITY ULTRASOUND IMAGING: CORRELATION BETWEEN NORMALIZED STRAIN AND RENAL CORTICAL FIBROSIS JING GAO,* WILLIAM WEITZEL,y JONATHAN M. RUBIN,z JAMES HAMILTON,x JUN LEE,{ DARSHANA DADHANIA,{ and ROBERT MIN* y
* Department of Radiology, New York-Presbyterian Hospital, Weill Cornell Medical College, New York, New York, USA; Department of International Medicine, University of Michigan Hospital and VA Medical Center, Ann Arbor, Michigan, USA; z Department of Radiology, University of Michigan Hospital, Ann Arbor, Michigan, USA; x Epsilon Imaging, Ann Arbor, Michigan, USA; and { Rogosin Institute, New York-Presbyterian Hospital, Weill Cornell Medical College, New York, New York, USA (Received 20 November 2012; revised 18 March 2013; in final form 11 April 2013)
Abstract—After transplantation, over a widely variable time course, the cortex of the transplanted kidney becomes stiffer as interstitial fibrosis develops and renal function declines. Elasticity ultrasound imaging (EUI) has been used to assess biomechanical properties of tissue that change in hardness as a result of pathologic damage. We prospectively assessed the hardness of the renal cortex in renal transplant allograft patients using a normalized ultrasound strain procedure measuring quasi-static deformation, which was correlated with the grade of renal cortical fibrosis. To determine cortical strain, we used 2-D speckle-tracking software (EchoInsight, Epsilon Imaging, Ann Arbor, MI, USA) to perform offline analysis of stored ultrasound loops capturing deformation of renal cortex and its adjacent soft tissue produced by pressure applied using the scanning transducer. Normalized strain is defined as the mean developed strain in the renal cortex divided by the overall mean strain measured in the soft tissues from the abdominal wall to pelvic muscles. Using the Banff scoring criteria for renal cortical fibrosis as the gold standard, we classified 20 renal transplant allograft biopsy tissue samples into two groups: group 1 (n 5 10) with mild (,25%) renal cortical fibrosis and group 2 (n 5 10) with moderate (26%–50%) renal cortical fibrosis. An unpaired two-tailed t-test was used to determine the statistical difference in strains between patients with mild and those with moderate renal cortical fibrosis. Receiver operating characteristic curve analysis was performed to assess the accuracy of developed strain and normalized strain in predicting moderate renal cortical fibrosis. The reference strain did not significantly differ between the two groups (p 5 0.10). However, the developed renal cortical strain in group 1 with mild fibrosis was higher than that in group 2 with moderate fibrosis (p 5 0.025). The normalized strain in group 1 was also higher than that in group 2 (p 5 0.0014). The areas under receiver operating characteristic curves for developed strain and normalized strain were 0.78 and 0.95, respectively. The optimal cutoff for distinguishing moderate renal cortical fibrosis was 20.08 for developed strain (sensitivity 5 0.50, specificity 5 1.0) and 2.5 for normalized strain (sensitivity 5 0.80, specificity 5 1.0). In summary, renal cortex strain is strongly correlated with grade of renal cortical fibrosis. Normalized strain is superior to developed strain in distinguishing moderate from mild renal cortical fibrosis. We conclude that free-hand real-time strain EUI may be useful in assessing the progression of cortical fibrosis in renal transplant allografts. Further prospective study using this method is warranted. (E-mail: [email protected]) Ó 2013 World Federation for Ultrasound in Medicine & Biology. Key Words: Elastography, Renal cortical fibrosis, Renal transplant, Ultrasound strain.
ratory tests and, if definitive diagnosis is needed, renal biopsy. However, routine laboratory tests such as serum creatinine are insensitive to loss of renal function (Hunsicker and Bennett 1995), and kidney biopsy is limited by its invasive nature and high cost (Schwarz et al. 2005). Therefore, improved non-invasive techniques to assess renal function and detect renal dysfunction early would be desirable. Elasticity ultrasound imaging (EUI) is an emerging non-invasive imaging
INTRODUCTION In current practice, assessment of renal transplant dysfunction relies on the monitoring of biochemical labo-
Address correspondence to: Jing Gao, Department of Radiology, New York-Presbyterian Hospital, Weill Cornell Medical College, 525 East 68th Street, Suite 8A, New York, NY 10065, USA. E-mail: [email protected] 1536
Renal transplant ultrasound elastography d J. GAO et al.
technique for the assessment of internal organ and tissue biomechanical properties that may be useful in assessing pathologic changes that influence tissue mechanics (O’Donnell et al. 1993; Xu et al. 2012). In the kidney, EUI may provide information about the mechanical changes that accompany histologic changes useful for monitoring renal disease (Chaturvedi et al. 1998). Quasi-static ultrasound strain measurements of the renal cortex in renal transplant recipients may help detect and monitor tissue hardness changes in a chronic transplant allograft nephropathy (CTAN) (Weitzel et al. 2004). More recently, techniques such as acoustic radiation force impulse imaging (Stock et al. 2011; Syversveen et al. 2011, 2012), shear wave velocity imaging (Gennisson et al. 2010; Ozkan et al. 2013) and shear wave dispersion vibrometry (Amador et al. 2009) have been used to assess renal cortical mechanical changes in renal transplants. Work remains to establish and improve the reliability and accuracy of these techniques, taking into account the effects of renal anisotropy and boundary conditions (Gennisson et al. 2012; Rubin et al. 1988; Syversveen et al. 2011). We chose to test and further evaluate the free-hand push method suggested in earlier work (Weitzel et al. 2004) because of (i) the locally isotropic renal cortical structure, wherein pathologic changes of interest develop relatively near the surface (transducer); (ii) the possibility of standardized data collection in the measurement by normalizing with a reference strain measurement from the acquired strain image; (iii) the availability of gold standard criteria for reference measurements using an accepted pathology scoring system for fibrosis in the transplant setting; and (iv) the simplicity of the measurement method, which would allow for broad clinical use if the method should prove useful. In this study we sought to prospectively assess, using EUI, the relationship between the hardness of renal cortex, represented by mean developed renal cortical strain measurements and normalized strain measurements (developed renal cortical strain divided by imaged soft tissue strain), and the grade of renal cortical fibrosis by Banff criteria. The ultimate goal of this study was to assess the value of ultrasound strain imaging in monitoring the progression of renal cortical fibrosis in the allograft after transplantation. METHODS Patients We performed EUI on 20 patients (11 men and 9 women, age range 28–87 y, mean age 50 6 16 y) who underwent transplanted kidney biopsy from March 2012 to August 2012. All patients were studied after providing written informed consent in this study. All color Doppler
1537
sonography and quasi-static elastography were conducted at the New York-Presbyterian Hospital, Weill Cornell Medical College, after obtaining approval from institutional review board of Weill Cornell Medical College. The study was compliant with the Health Insurance Portability and Accountability Act. Patients were eligible to participate if a referring nephrologist or transplant surgeon referred them for renal biopsy. The indications for renal biopsy were loss of renal function, suspicion of allograft rejection, or participation in protocol renal transplant biopsy. As a standard of care for patients having renal transplant in our institution, color Doppler sonography was routinely performed to screen for allograft vascular complications, for example, transplant renal artery stenosis, and non-vascular complications, such as hydronephrosis for an elevated serum creatinine. Patients with hydronephrosis, large perinephric collections, immediate post-transplantation status (,7 d after the surgery), significant transplant renal artery stenosis and an existing intrarenal vascular abnormality (e.g., arteriovenous fistula), which may be contraindications for renal biopsy and/or affect renal cortical strain measurement, were not enrolled in the study; in addition, patients who could not tolerate compression for any reason were excluded. Real-time ultrasound data acquisition Two investigators who had more than 20 y of experience in ultrasound scanning and were trained in freehand compression performed the scans (one investigator scanned 15 cases, and the other scanned 5 cases). Gentle but firm free-hand compression of a transplanted kidney was performed during standard renal transplant sonography before kidney biopsy, on the same day as the biopsy (n 5 17) or ,7 d before the biopsy (n 5 3). Subjects were placed in the supine position. The transplanted kidney was imaged using either a General Electric (Logic E9, General Electric, Milwaukee, WI, USA) or Siemens (Sequoia 512, Siemens Medical Solution, Mountain View, CA, USA) scanner equipped with a multi-frequency 2- to 4-MHz curved linear array transducer. Transmission gel was placed on the anterior abdominal wall as standard acoustic coupling for ultrasound examination over the region where the transplanted kidney is located. The speckle reduction setting on the Logic E9 scanner was turned off while capturing realtime compression to increase the fidelity of speckle data acquisition. The speckle reduction feature performs an averaging function on the images, making the examination more pleasing to the eye, but may influence tracking and therefore was disabled as part of this protocol. We used a transducer frequency range of 2–4 MHz to improve penetration of the ultrasound beam to the deep pelvis. The depth of gray-scale imaging ranged
1538
Ultrasound in Medicine and Biology
Volume 39, Number 9, 2013
Fig. 1. (a) Real-time gray-scale image of free-hand compression on the short axis of a transplanted kidney recorded prebiopsy. The image was from a case in group 1 (,25% renal cortical fibrosis), a patient who had undergone transplantation of a living non-related donor kidney 4 mo earlier. Regional strain measurements were performed using connected dual regions of interest (ROIs), which are axially placed with respect to the center of the transducer (cyan and red dotted lines). The movement of each ROI of the tethered pair through the image loop is calculated from tissue motion estimates produced by speckle tracking. Developed strain (the line between the two cyan arrows) is the strain in renal cortex, and reference strain (the line between the two red arrows) is the strain in the soft tissue from the anterior abdominal wall to pelvic muscles. (b) Graph of the strain measurements from (a). Developed strain (cyan solid line) is significantly higher than reference strain (red solid line). Normalized strain 5 developed strain/reference strain. In this case, developed strain, the reference strain and normalized strain are 20.10, 20.02 and 5, respectively. Histopathologic analysis of the kidney biopsy specimen revealed ,5% interstitial fibrosis, activated lymphocytes, marked active tubulitis and interstitial inflammation.
from 9 to 13 cm, as a transplanted kidney is usually located superficially in the right or left lower quadrant just under the anterior abdominal wall. We used conventional gray-scale imaging to select the region from the skin of the anterior abdominal wall to pelvic muscles that was free of bowel and urinary bladder after routinely imaging a transplanted kidney. The conventional ultrasound examination included gray-scale, color Doppler and spectral Doppler analysis. We started an applied axial compression (Kallel et al. 1997) by pressing on the anterior abdominal wall and underlying tissue including the transplanted kidney and pelvic muscles. The pushing force applied was similar to deep palpation in a physical examination, but the ultrasound transducer was used while monitoring the subject to ensure his or her comfort during the ultrasound palpation examination. The pushing force was directed toward the iliac, and the deformation was measured down to the iliopsoas muscle in the pelvis. A compression was considered effective when the deformation in the region of interest reached 10% of the initial depth of soft tissue (Emelianov et al. 1995; Xu et al. 2012) measured from the abdominal wall to the iliopsoas muscle. Compression was conducted in sagittal and/or transverse views of the transplanted kidney. On conclusion of the application of constant pressure on the kidney, the pressure was quickly released while ultrasound data capture continued to provide pre- and
post-push reference images for post-procedure data processing and speckle tracking (Emelianov et al. 1998). To improve the accuracy and robustness of strain measurements, the acquisition frame rate was increased to reduce speckle de-correlation between frames. Through use of a single transmit focus and reduced spatial resolution (i.e., beam density), frame rates of 33–47 Hz were achieved for full field-of-view imaging covering the longitudinal or transverse section of the kidney. In addition, the ultrasound probe was carefully controlled to apply compression normal to the transducer face, further limiting speckle de-correlation by minimizing out-of-plane tissue motion. All captured images were stored in the Picture Achieving and Communications System (PACS) in the Department of Radiology. 2-D speckle tracking for measuring renal transplant strains Gray-scale image loops were captured and stored in Digital Imaging and Communications in Medicine (DICOM) format and pre-processed by speckle tracking software (EchoInsight, Epsilon Imaging, Ann Arbor, MI, USA) to prepare for strain analysis. For this study, regional strain measurements were done using a connected, dual region of interest (ROI), as shown by the cyan- and red-colored graphics in Figures 1(a, b) and 2
Renal transplant ultrasound elastography d J. GAO et al.
1539
Fig. 2. (a) Image of free-hand compression on a longitudinal section of a transplanted kidney from a case in group 2 (26%–50% renal cortical fibrosis), who had an elevated creatinine level (3.2 mg/dL) in his fourth year after renal transplantation. Strain measurements were performed along axial of the cortex at the center of the transducer (cyan dotted line and red dotted line). Care should be taken to avoid calyces and renal pelvis (red arrow) and medulla pyramid (white arrow), which may introduce error into measurements of developed strain. (b) Graph of the strain measurements from (a). In this case, developed strain (cyan line), reference strain (red line) and normalized strain (developed strain/reference strain) are 20.10, 20.08 and 1.25, respectively. Histopatholic analysis of the kidney biopsy specimen revealed 30% interstitial fibrosis, active moderate tubular atrophy and chronic active antibodymediated rejection. It is clear from this graph that the difference between developed strain and reference strain is not significantly different in this case of moderate renal cortical fibrosis. Compared with the case of mild renal cortical fibrosis in Figure 1, although using a relatively higher reference strain (20.08 vs. 20.02), normalized strain in this case of moderate renal cortical fibrosis is significantly lower than that in mild renal cortical fibrosis (1.25 vs. 5).
(a, b). The movement of each ROI of the tethered pair through the image loop is calculated from tissue motion estimates produced by speckle tracking. The relative motion between the ROIs is expressed as strain, and represents the average tissue deformation between the ROI pair (i.e., total strain along the connecting line of the ROI tether). In this study, developed strain is defined as the strain in the renal cortex between the renal capsule and the collecting system (range, 0.8–1.5 cm in anteriorposterior dimension, shown as a cyan line in Fig. 1a, b), and reference strain is total deformation from the skin of the abdominal wall to the pelvic muscles after compression (shown as a red line in Fig. 1a, b). Care should be taken to avoid capsule, calyx, medulla and sinus fat that may cause an error in measuring developed strain (Fig. 2a). Preferably, strain is measured along the direction of compression, in this case the normal from the transducer surface at the center of the image (i.e., vertical direction in Fig. 2a). The developed strain (cyan) and reference region (red) during compression are indicated in Figures 1b and 2b. All strain measurements for this study were made by a single operator (J.G.). For each subject, both developed strain and reference strain were measured using these methods.
Kidney biopsy and histopathology Ultrasound-guided transplanted kidney biopsy was performed by the nephrologist using our standard protocol. Patients fasted 8–10 h before the kidney biopsy. The procedure was performed with patients in the supine position. A Siemens scanner equipped with a 4V1 sector array transducer (Sequoia 512, Siemens Medical Solution) with biopsy guide (Ultra-Pro II Needle Guide, CIVCO Medical Instruments, Kalona, IA, USA) in a sterile transducer cover provided real-time, gray-scale images to guide local anesthesia and kidney biopsy. The mid- to lower pole of the kidney was commonly selected for kidney biopsy, and an area lacking prominent vessels on sonography was chosen as the biopsy site. To obtain an adequate biopsy specimen and avoid damaging renal vessels, the angle between the biopsy needle and the skin was 18 or 32 depending on the thickness of the renal cortex and existing vessels along the biopsy path. An 18-gauge biopsy needle (Bard Peripheral Technologies, Covington, GA, USA) was inserted and advanced to the kidney cortex under ultrasound guidance after local anesthesia. Static images and cine loops of renal sonography and biopsy were stored in the PACS system in the Department of Radiology.
1540
Ultrasound in Medicine and Biology
Volume 39, Number 9, 2013
Table 1. Clinical information and ultrasound strain on 20 renal transplants
Male/female ratio Live donor/deceased donor Age Duration of transplant (mo) Creatinine (ng/dL) Glomerular filtration rate Reference strain Developed strain Normalized strain*
Group 1 (,25% cortical fibrosis)
Group 2 (36%–50% cortical fibrosis)
p (t-test)
3/7 7/3 46 6 16 21.3 6 27.7 1.89 6 1.16 42.81 6 6.8 20.03 6 0.02 20.14 6 0.05 5.23 6 2.64
8/2 1/9 58 6 12 46.2 6 37.1 3.49 6 0.81 18.8 6 6.48 20.05 6 0.09 20.08 6 0.04 1.73 6 0.83
0.07 0.065 0.002 0.0005 0.100 0.025 0.0014
* Normalized strain 5 developed strain/reference strain.
A pathologist with more than 25 y of experience in the interpretation of transplanted kidney biopsies reviewed biopsy specimens; she was blinded to the results of renal sonography. Biopsies were evaluated on the basis of the Banff’07 classification of renal allograft pathology (Solez et al. 2008). Further, the glomeruli, tubules, interstitium and vessels were evaluated. Special stains used were periodic acid-Schiff, periodic acid-silver methenamine and trichrome. The severity of interstitial fibrosis/tubular atrophy was estimated to the nearest 5% and then divided into four categories: 0, no fibrosis present; 1, mild (,25%) fibrosis; 2, moderate (26%–50%) fibrosis; and 3, severe (.50%) fibrosis.
The difference in reference strain between the two groups did not reach statistical significance (p 5 0.10), and of note, this reference strain was lower in the lowfibrosis group than in the high-fibrosis group. The differences in developed strain and normalized strain between the two groups were statistically significant. Developed strain in group 1 with mild renal cortical fibrosis was higher than that in group 2 with moderate renal cortical fibrosis (p 5 0.025). Developed cortical strain was greater in group 1 even though the overall (background) strain was lower in this group compared with group 2. Further, normalized strain was significantly higher in group 1 than in group 2 (p 5 0.0014). The area under the ROC curve for developed strain was 0.78 (Fig. 3), whereas that for normalized strain was 0.95 (Fig. 4).
Statistical analysis All variables including developed strain (i.e., strain in renal cortex), reference strain and normalized strain are expressed as means and standard deviations (SD). An unpaired two-tailed t-test (Microsoft Excel) was used to analyze the difference in strain between group 1 with ,25% renal cortical fibrosis and group 2 with 26%–50% renal cortical fibrosis based on Banff cortical fibrosis score criteria. A p value ,0.05 was considered statistically significant. Diagnostic sensitivity and specificity were evaluated on the basis of the receiver operating characteristic curve (ROC) analysis (www.rad.jhmi.edu/ jeng/javarad/roc). The optimal cutoff values were chosen to maximize the sum of sensitivity and specificity. RESULTS Characteristics, including age, gender, duration of transplant, donor status and renal function tests, of 20 patients are listed in Table 1. There were no significant differences in patient age or duration of transplant between the two groups. There were significant differences in serum creatinine (p , 0.002) and glomerular filtration rate (GFR) (p , 0.0005) between the two groups.
Fig. 3. Receiver operating characteristic (ROC) curve analysis was performed to assess the accuracy of developed strain in predicting moderate renal cortical fibrosis. The area under the curve is 0.78. For developed strain, the optimal cutoff for distinguishing moderate renal cortical fibrosis is 20.08 (sensitivity 5 0.50 and specificity 5 1.0).
Renal transplant ultrasound elastography d J. GAO et al.
Fig. 4. Receiver operating characteristic (ROC) curve analysis was performed to assess the accuracy of normalized strain in predicting moderate renal cortical fibrosis. The area under the curve is 0.95. For normalized strain, the optimal cutoff for distinguishing moderate renal cortical fibrosis is 2.5 (sensitivity 5 0.80 and specificity 5 1.0).
Normalized strain appears to have better diagnostic value than developed strain (0.95 vs. 0.78) in moderate renal cortical fibrosis. The optimal cutoff values for detecting moderate renal cortical fibrosis were 20.08 (sensitivity 50% and specificity 100%) and 2.5 (sensitivity 80% and specificity 100%) for developed strain and normalized strain, respectively (Table 2).
DISCUSSION In quasi-static elastography, echo signals acquired before and after a small applied deformation (about 1% inter-frame compression) are correlated to estimate tissue displacements (Varghese 2009). The resulting local tissue displacements are used to estimate deformation by calcuTable 2. Receiver operating characteristic curves (one observer)
Area under receiver operating characteristic curve Cutoff value* Sensitivity Specificity Positive predictive value Negative predictive value 95% Confidence interval Significance level, p
Developed strain
Normalized strain
0.78
0.95
20.08 50% 100% 100% 66.7% 51%–91% 0.025
2.5 80% 100% 100% 88.3% 75%–99% 0.0014
* Cutoff value is applied to distinguish moderate from mild renal cortical fibrosis.
1541
lating strain gradients from the tissue motion field. EUI techniques are based on the hypothesis that soft tissues deform more than stiffer tissues, providing an estimate of tissue stiffness. In free-hand EUI, even if the applied surface stress is kept constant, the subsurface stress may vary from patient to patient and from one examination to another in the same patient. When the subsurface stress is uncertain, methods need to be derived to fairly compare the resulting strain. For any EUI technique to be clinically useful, and in the absence of a full knowledge of the conditions needed to determine Young’s modulus, a simple method of gauging the relative stress (force) between examinations is needed. We chose to normalize the developed strain (deformation) in the renal cortex by the background strain as described under Methods. Although not fully accounting for many variables that influence strain, this method of normalizing strain did increase the sensitivity and specificity of the strain measurement in distinguishing moderate from mild renal cortical fibrosis in our population. There are other methods of strain normalization that may be effective (Weitzel et al. 2004). Although current results suggest that normalization is important to improve the diagnostic accuracy of the measurement, the optimal method of normalization remains to be determined. Previously published methods of normalization based on the entire cross section of the imaged kidney (Weitzel et al. 2004) may have advantages in determining the strain relative to the overall strain of the kidney; however, such methods require the entire kidney to be in the ultrasound field of view throughout the compression measurement. The normalized cortical strain in this study has the practical advantage that only the region of interest (transplanted kidney cortex) and region of reference (soft tissue from the abdominal wall to pelvic muscles) must remain in the field of view during compression. It should be noted that may influence measurement beyond the linearity and normalizing assumptions we applied in this study. For example, tissues are inherently non-linear (Emelianov et al. 1995; Xu et al. 2012) in their stress-strain characteristics. For high static strain, there is reduced incremental strain with additional applied stress as compared with low static strain. This non-linear stressstrain behavior may decrease the diagnostic discrimination of the test. Concurrently, increased strain may at the same time lead to patient discomfort. Conversely, if too little pressure is applied, there may not be enough strain developed in either group to make meaningful measurements. We observed mean strain values of 20.109 that both kept patients perfectly comfortable during the examination and still allowed discrimination between degrees of fibrosis. Although these results were achieved with an examination by an experienced
1542
Ultrasound in Medicine and Biology
operator in this study, standardization of the relationship between developed strain to applied stress and standardized strain targets during the examination may allow for even greater combined sensitivity and specificity, while minimizing or accounting for non-linear effects and maintaining a safe and comfortable EUI examination in future studies and, eventually, in clinical practice. There are limitations to this study. (i) The number of cases enrolled in the study was small. (ii) Tests of intraand inter-observer variation need to be investigated, as only one operator performed only one measurement of strain in this study. (iii) Different factors could potentially affect the accuracy of renal strain measurements, including changes in depth of the ROI and geometry of acquisition, that is, curved array versus linear array transducers. Such changes require further assessment. (iv) Pathologic changes in renal allografts are generally complex, in which renal cortical fibrosis is a marker of prognosis after transplantation. Renal cortical fibrosis may have been a major cause of transplant dysfunction at the time of data collection in this study. However, there are other factors that could result in renal transplant dysfunction such as cellular rejection, mediated rejection and nephrotoxicity. The correlations between ultrasound strain and those pathologic factors require further investigation. In summary, the hardness of the renal cortex, measured as developed strain and normalized strain using the EUI method with ultrasound palpation, is strongly correlated with the degree of renal cortical fibrosis. This method or similar methods may prove to be a valuable tool in the evaluation of renal transplant patients. More work is needed using these or other standardized ultrasound strain data acquisition methods to perform prospective clinical studies to determine the optimal role of EUI in managing renal transplant recipients. REFERENCES Amador C, Urban MW, Warner LV, Greenleaf JF. In vitro renal cortex elasticity and viscocity measurements with shearwave dispersion ultrasound vibrometry (DSUV) on swine kidney. Conf Proc IEEE Eng Med Biol Soc 2009;2009:4428–4431. Chaturvedi P, Insana MF, Hall TJ. Ultrasouns and elasticity imaging to model disease-induced changes in soft-tissue structure. Med Image Anal 1998;2:325–338. Emelianov SY, Erkamp RQ, Lubinski MA, Skovoroda AR, O’Donnell M. Non-linear tissue elasticity: Adaptive elasticity imaging for large
Volume 39, Number 9, 2013 deformations. In: Proceedings, 1998 IEEE Ultrasonics Symposium, Sendai, Japan, October 5–8, 1998:1753–1756. Emelianov SY, Lubinski MA, Weitzel WF, Wiggins RC, Skovoroda AR, O’Donnell M. Elasticity imaging for early detection of renal pathology. Ultrasound Med Biol 1995;21:871–883. Gennisson JL, Grenier N, Hubrecht R, Couzy L, Delmas Y, Derieppe M, Lepreaux S, Merville P, Criton A, Bercoff J, Tanter M. Multiwave technology introducing shear wave elastography of the kidney: Pre-clinical study on a kidney fibrosis model and clinical feasibility study on 29 human renal transplants. In: Proceedings, 2010 IEEE Ultrasonics Symposium October 11–14, 2010:1356–1359. Gennisson JL, Grenier N, Combe C, Tanter M. Supersonic shear wave elastography of in vivo pig kidney: Influence of blood pressure, urinary pressure and tissue anisotropy. Ultrasound Med Biol 2012; 38:1559–1567. Hunsicker LG, Bennett LE. Design of trials of methods to reduce late renal allograft loss: The price of success. Kidney Int Suppl 1995; 52:S120–S123. Kallel F, Ophir J, Maguee K, Krouskop T. Quantitative imaging of lowcontrast elastic modulus distribution in tissues using elastography. In: Proceedings, 1997 IEEE Ultrasonics Symposium, Toronto, ON, October 5–8, 1997:1439–1442. O’Donnell M, Emelianov SY, Skovoroda AR, Lubinski MA, Weitzel WF, Wiggins RC. Quantitative elasticity imaging. In: Proceedings, 1993 IEEE Ultrasound Symposium, Baltimore, MD, October 31–November 3, 1993:893–902. Ozkan F, Yavuz YC, Inci MF, Altunoluk B, Ozcan N, Yuksel M, Sayarlioglu H, Dogan E. Interobserver variability of ultrasound elastography in transplant kidneys: Correlations with clinical-Doppler parameters. Ultrasound Med Biol 2013;39:4–9. Rubin JM, Carson PL, Meyer CR. Anisotropic ultrasonic backscatter from the renal cortex. Ultrasound Med Biol 1988;14:507–511. Schwarz A, Gwinner W, Hiss M, Radermacher J, Mengel M, Haller H. Safety and adequacy of renal transplant protocol biopsies. Am J Transplant 2005;5:1992–1996. Solez K, Colvin RB, Racusen LC, Haas M, Sis B, Mengel M, et al. Banff ’07 classification of renal allograft pathology: Updates and future directions. Am J Transplant 2008;8:753–760. Stock EF, Klein BS, Vo Cong MT, Regenbogen C, Kemmner S, Buttner M, Wagenpfei S, Matevossian E, Renders L, Heemann U, Kuchle C. ARFI-based tissue elasticity quantification and kidney graft dysfunction: First clinical experiences. Clin Hemorheol Microcirc 2011;49:527–535. Syversveen T, Brabrand K, Midtvedt K, Strom EH, Hartmann A, Jakobsen JA, Berstad AE. Assessment of renal allograft fibrosis by acoustic radiation force impulse quantification: A pilot study. Transplant Int 2011;24:100–105. Syversveen T, Midtvedt K, Berstad AE, Brabrand K, Strom EH, Abildgaard A. Tissue elasticity estimated by acoustic radiation force impulse quantification depends on the applied transducer force: An experimental study in kidney transplant patients. Eur Radiol 2012; 22:2130–2137. Varghese T. Quasi-static ultrasound elastography. Ultrasound Clin 2009; 4:323–338. Weitzel WF, Kim K, Rubin JM, Wiggins RG, Xie H, Chen X, Emelianov SY, O’Donnell M. Feasibility of applying ultrasound strain imaging to detect renal transplant chronic allograft nephropathy. Kidney Int 2004;65:733–736. Xu J, Tripathy S, Rubin JM, Stidham RW, Johnson LA, Higgins PD, Kim K. A new nonlinear parameter in the developed strain-toapplied strain of the soft tissue and its application in ultrasound elasticity imaging. Ultrasound Med Biol 2012;38:511–523.