First published online September 13, 2006
British Journal of Radiology (2006) 79, 935-942
© 2006 British Institute of Radiology
doi: 10.1259/bjr/53140218
A simple processing method allowing comparison of renal enhancing volumes derived from standard portal venous phase contrast-enhanced multidetector CT images to derive a CT estimate of differential renal function with equivalent results to nuclear medicine quantification
J C Fowler, MA, FRCR, MSc
C Beadsmoore, MBBS, MRCP, FRCR
M T G Gaskarth, MA, FRCS, FRCR
H K Cheow, MBChB, FRCR, MSc
R Bernal, MD, MSc
P Hegarty, FRCS (Urol), MCh
K N Bullock, MD, FRCS
H Taylor, MA, MRCP, FRCR
A K Dixon, MD, FRCR, FMedSci
and
A M Peters, MD, FRCR, FMedSci
Departments of Radiology, Nuclear Medicine and Urology, Addenbrooke's Hospital and the University of Cambridge, Cambridge, UK
Correspondence: Dr J Charlotte Fowler, Department of Diagnostic Imaging, Luton and Dunstable NHS Trust, Luton LU4 0DZ, UK. E-mail: charlotte.fowler{at}ldh.nhs.uk
 |
Abstract
|
|---|
As iodinated contrast medium is cleared by glomerular filtration, it should be possible to apply the same principles utilized in radionuclide studies to derive differential renal function by comparison of enhancing renal volumes derived from contrast enhanced multidetector CT (CEMDCT). Having established a technique iteratively which appeared successful, a retrospective study was performed using 25 consecutive patients with a wide range of urological conditions who had undergone both CEMDCT, including the renal area in the portal venous phase, and nuclear medicine (NM) assessment of renal function with no urological intervention between the studies. Proprietary volume software was used to quantify the volume and attenuation of each kidney, the products of which (after subtraction of soft tissue attenuation derived from a region of interest over psoas) gave right and left enhancing renal volumes. The contribution by each kidney as a percentage of total renal enhancing tissue was derived. Comparison with NM studies resulted in excellent correlation of relative renal function by CEMDCT and NM assessments having a regression of near unity and a Pearson's correlation coefficient of 0.96. Bland Altman and Passing Bablock tests confirmed good agreement between the two methods with no bias. This is a simple, practicable processing technique using standard portal venous phase CEMDCT images to quantify differential function. This technique may allow a one-stop CT assessment of both anatomy and function.
|
Introduction
|
|---|
Patients undergoing investigation for urological problems can have two diagnostic imaging studies: contrast enhanced multidetector CT (CEMDCT) for anatomy and a nuclear medicine (NM) study for estimation of the contribution by each kidney to global renal function, referred to henceforth as the differential renal function (DRF).
Iodinated contrast media is cleared by glomerular filtration. There is a linear relationship between contrast media concentration and attenuation. It should therefore be possible to derive DRF from iodinated contrast medium enhanced CT using similar methods of analysis to those used in NM renography by comparing enhancing volumes of the kidney prior to onset of drainage of urinary contrast media. Previous studies have demonstrated that this is possible, but the techniques described required either excessive additional radiation [1–4] or considerable operator input [5–9] to be useful in routine clinical practice. The aim of this study was to establish whether a simple data processing technique could be applied to enhancing volumes of kidneys from CEMDCT imaged in the portal venous phase to obtain a useful CT derived DRF using NM derived DRF as the gold standard.
There are three commonly used radioisotope tracers available for assessment of renal function, 99Tcm-diethylenetriaminepentaacetic acid (DTPA), 99Tcm-dimercaptosuccinic acid (DMSA) and 99Tcm-mercaptoacetyltriglycine (MAG3). These have different methods of renal handling, DTPA being cleared by glomerular filtration, DMSA retained by active renal tubular uptake and MAG3 cleared almost exclusively by renal tubular secretion. Despite these different methods of renal handling, the three agents give similar DRF and in clinical practice are used interchangeably for this purpose. DTPA was not used in this study as it has inferior dosimetry and background clearance compared with MAG3. It was therefore considered reasonable to compare CEMDCT-derived DRF with NM results derived from DMSA and MAG3 studies.
|
Methods
|
|---|
A post-processing technique was developed iteratively using CEMDCT renal images of a patient who had a contemporaneous DMSA study, which was used as the gold standard.
The CEMDCT images had been acquired using a standard portal venous phase protocol on a Siemens Sensation (Forchheim, DE) 16 slice multidetector CT system (rotational time 0.5 s, detector collimation 16 mm x1.5 mm, table feed per rotation 18 mm) resulting in an abdomen/pelvis study being acquired in 12 s, with imaging commencing at 90 s following a 100 ml pump injection of iopamidol 300 at a rate of 2.5 ml s–1.
Variations to the technique were made, applying the principles of NM processing, until a simple and robust method was achieved that resulted in an equivalent DRF derived by both CEMDCT and NM methods as follows:
The Siemens proprietary Leonardo volume software was used to obtain left and right renal volumes of interest (VOIs) by swift placement of area regions of interest (ROIs) around the kidneys (including some surrounding perinephric and pelvicalyceal fat, and pelvicalyceal systems) on every 3rd or 4th axial image. The software automatically interpolated the intervening axial images to allow the whole renal volume to be selected. It also allowed the selection of window levels that were used to exclude non-enhancing structures within the VOI. Pelvicalyceal and perirenal fat, cysts, and hydronephrotic pelvicalyceal fluid and poorly enhancing tumours were excluded by applying a lower window limit of 20 HU. Stones were excluded by applying an upper window limit of 400 HU. Information provided by the software included volume (V) in cm3 and mean Hounsfield Units (HU) for each VOI (Figure 1
).
View larger version (97K):
[in this window]
[in a new window]
|
Figure 1. Snapshot of Leonardo volume analysis software showing right(orange) and left (pink) renal volume of interest (VOI). Windowing has been applied to exclude perirenal and pelvicalyceal fat.
|
|
An estimation of non-enhanced renal tissue attenuation (NEA) was derived from a ROI over the adjacent psoas muscles that could be subtracted from the measured renal attenuation to derive the enhancing attenuation for each kidney. For the right kidney, the enhancing renal volume (RERV) was derived as follows: RERV = RVx(RHU–NEA). The same calculation was made for the left. The CEMDCT derived DRF for the right kidney was therefore [RERV/( RERV+ LERV)]x100.
Having developed this method iteratively it was then tested on a wider population. The radiology information system was used to generate a list of consecutive patients who had undergone CEMDCT in the portal venous phase. Cross referencing these patients with the NM database and clinical information from investigation request forms allowed identification of patients who had had both an appropriate CEMDCT study and NM assessment of DRF by either DMSA or MAG3 within a 2 month period. Patients were excluded if they had undergone urological intervention of any type during the intervening period. 25 patients matched the criteria. Operators blinded to the NM derived DRF result used the described CT processing method to establish the CEMDCT-derived DRF, which were compared with the NM-derived DRF for each patient. Unenhanced renal volume data from MDCT was used to establish volume derived DRF for each patient.
Statistical analysis
Statistical analysis was performed using MedCalc® software (MedCalc Software, Mariakerke, Belgium). Pearson's correlation coefficient and two methods of measurement comparison, Bland Altman and Passing Bablock tests, were used to establish whether the results from CEMDCT and NM were statistically equivalent.
|
Results
|
|---|
Table 1 details the patients in terms of age, sex, time between CEMDCT and NM studies, indications for study, CT and histopathology results and DRF derived from NM, CEMDCT and unenhanced MDCT volume. The patient's ages ranged from 16 years to 85 years. There was an equal split between males and females. There was an average of 15 days between CEMDCT and their NM studies. There was a wide range of indications for their studies including renal cell carcinoma (4), transitional cell carcinoma (5), Von Hippel-Lindau related clear cell adenocarcinoma (1), schistosomiasis related squamous cell carcinoma (1), undifferentiated sarcoma (1), renal colic (3), renal artery stenosis (1), retroperitoneal fibrosis (1), chronic pelviureteric junction obstruction (2), unexplained hydronephrosis (2), cystitis (1), pain (2) and trauma (2). There was generally very good correlation of CEMDCT and NM derived DRF, ranging from 0 to 8% with a mean of 3% difference between the two modalities. There was poorer correlation between unenhanced volume and NM derived DRF, ranging from 2% to 50%, with a mean of 9% difference between the two modalities.
View this table:
[in this window]
[in a new window]
|
Table 1. Patient details including age, sex, clinical condition, DRF derived from NM, CEMDCT and unenhanced MDCT renal volumes
|
|
Figure 2a compares CEMDCT and NM-derived DRF. There is a linear relationship between DRF derived by CEMDCT and NM, the gradient of which is close to unity (y = 1.05x–2.5). There was very good correlation (r = 0.96; n = 26; p<0.0001). The Bland Altman plot (Figure 3) shows no bias with change in magnitude and the Passing Bablock analysis confirms no significant deviation from linearity (p>0.1). Figure 2b compares unenhanced volume from MDCT and NM-derived DRF. There was no linear relationship (y = 0.02x+52) and poor correlation (R2 = 0.02).
View larger version (21K):
[in this window]
[in a new window]
|
Figure 2. (a) Correlation of contrast enhanced multidetector CT (CEMDCT) and nuclear medicine (NM)-derived differential renal function (DRF) with respect to the contribution by the left kidney (n = 25). (b) Correlation of non-enhanced volume from MDCT and NM-derived DRF with respect to the contribution by the left kidney (n = 25).
|
|
View larger version (13K):
[in this window]
[in a new window]
|
Figure 3. Bland Altman plot of contrast enhanced multidetector CT(CEMDCT) and nuclear medicine (NM)-derived differential renal function (DRF) with respect to the contribution by the left kidney.
|
|
|
Discussion
|
|---|
The renal time-concentration curve of an indicator filtered at the glomerulus has a characteristic shape that is a manifestation of the flow of tracer through the nephron following a small volume bolus intravenous injection (Figure 4). The first peak (phase 1) is due to perfusion as blood enters the kidney. This is followed by an uptake phase (phase 2) as filtered indicator enters the glomeruli and starts to travel along the nephrons. Once indicator starts to enter the collecting system and drain from the kidney, the curve starts to flatten. This occurs at the minimum parenchymal transit time (MPTT) from about 2.5 min depending on the patient's state of hydration and diuresis. The curve then reaches a plateau when the amount of indicator entering the glomeruli is equivalent to that leaving the system, after which time the gradient starts to decline as draining indicator exceeds incoming (phase 3). Iodinated contrast medium is cleared by glomerular filtration and follows the same pattern, familiar to radiologists (albeit in a snapshot series) during intravenous urography. The longer injections used in contrasted CT studies would result in an elongated renogram, but the MPTT of the kidney would remain unchanged as the first molecules of contrast medium pass out of the collecting system.
View larger version (11K):
[in this window]
[in a new window]
|
Figure 4. Renal time-concentration curve of a small volume bolus indicator injection cleared by glomerular filtration indicating phases 1 (perfusion phase), 2 (uptake phase, terminating at the minimum parenchymal transit time (MPTT)) and 3 (drainage phase).
|
|
Two nuclear medicine processing methods are commonly used to quantify the DRF from the renogram curves generated from activity within the two renal regions of interest following correction for background activity. The first, the Rutland Patlak method, compares the gradients of the right and left phase 2 uptake following integration of the arterial input, while the other compares the area under the curve (AUC) up to the end of phase 2 (i.e. prior to MPTT). The concepts behind these two analysis methods can be applied to CECT images as long as the precise time of image acquisition from the commencement of injection is known.
CECT-derived renal glomerular filtration rate (GFR) has been shown to be successful in the research setting using the Rutland Patlak gradient analysis with repeated data acquisitions through a single slice to derive the gradients [1–4]. However, such repeated acquisitions result in a high radiation exposure and these methods have not been reported in clinical practice.
The alternative method of comparison of AUC can be applied to CT (albeit as a snapshot at the time of image acquisition, which reflects height of curve rather than area) with comparison of enhancing volume of kidneys prior to MPTT. This method has been used previously and shown to be broadly successful in comparison with NM quantification [5–9]. These studies used enhancing volume data obtained by manual selection by precise ROI drawing around enhancing renal tissue on consecutive slices to exclude adjacent fat and other non-renal parenchymal tissues and the manual addition of axial data to build up volumes, a time consuming process susceptible to introduction of operator error and unlikely to be practicable in routine clinical practice. Earlier studies [5, 8] were also hampered by the use of slower conventional or single slice spiral CT, which could introduce artefact especially in patients with kidneys at significantly different craniocaudal positions within the body, resulting in their kidneys being imaged in different phases of contrast medium handling, with the most cranial kidney being imaged in phase 2 whereas the lower kidney may have reached phase 3 by the time image acquisition was carried out [5, 6].
The data processing method described in the current study using portal venous CEMDCT with subsecond data acquisition resulting in the whole of both kidneys being imaged at the same phase of renal enhancement. Imaging at 90 s following a 100 ml pump injection of iodinated contrast medium at a rate of 2.5 ml s was ideal for CEMDCT-derived DRF as data were acquired prior to MPTT in the late nephrographic phase. Use of the volume analysis software allowed swift drawing of ROIs on non-consecutive slices, which the software then interpolated for the whole renal volume. Use of windowing levels automatically excluded perirenal and pelvicalyceal fat, fluid in cysts or distended pelvicalyceal systems, poorly enhancing tumours and stones, ensuring only enhancing renal parenchymal VOI were included in the analysis with a minimum of operator effort. Single figures for the volume and mean HU for each kidney were obtained which, following correction for unenhanced renal soft tissue using a psoas ROI surrogate, resulted in the CEMDCT derived DRF, produced without excessive operator time, taking around 5 min.
A recent article [9] using the same scanner and similar image acquisition parameters to ours showed promising results comparing CEMDCT renal volumes with NM in patients with renal artery stenosis. Their methods resulted in similar close correlation (r2 = 0.93) between the CEMDCT and NM derived DRF as in the current study, but their results did not fit a line of unity, resulting in an equation of y = 0.75x+13.9 for CEMDCT derived DRF vs NM derived DRF unlike ours, which was close to unity with an intercept near zero (y = 1.05x–2.5). Therefore, in contrast to the current study, Björkman's method would not result in DRF data which could be directly interchanged with NM derived quantification. Björkman points out that this is likely to have arisen because they did not correct for soft tissue attenuation before comparing the enhancing volumes and our results, which includes this correction, supports this opinion. Their image acquisition during the arterial phase was likely to have compounded this further.
Another difference between the technique described by Björkman and our method is that despite using the Leonardo software for manual drawing of ROIs to establish renal VOI by addition of consecutive slices, they did not use the Leonardo volume software to derive renal volume data, instead using the PACS VOXAR software without the use of windowing to exclude non-enhancing renal tissue from the VOI, which resulted in widely differing results. Thus, despite having the technology available, they fell back on time consuming manual processing methods of excluding non-enhancing renal tissue.
Some authors have claimed that non-enhanced renal volume can be used to estimate DRF [10]. We have found that non-enhanced renal volumes do not correlate with NM derived DRF; our 23rd case involving renal infarction following trauma demonstrates this well (Figure 5). The CEMDCT derived DRF and NM derived DRF data were in agreement that the right kidney was non-functioning, whereas volume based data alone resulted in incorrect estimation of similar function from both kidneys.
View larger version (62K):
[in this window]
[in a new window]
|
Figure 5. (a) Contrast enhanced multidetector CT (CEMDCT) coronal reformat of patient 23 showing lack of enhancement of the right kidney. (b) 99Tcm-dimercaptosuccinic acid (DMSA) image of patient 23. CEMDCT derived differential renal function (DRF) agreed with nuclear medicine (NM) result that the right kidney was non-functioning. Unenhanced CT volume could not have been used as a method of deriving NM equivalent DRF.
|
|
The current method appears to be applicable across a wide range of urological pathologies, but there are a few situations where caution is required. First is that of the enhancing renal cell carcinoma (e.g. case 12, Figure 6) where the enhancement of the tumour may not be differentiated from that of the renal tissue by the windowing process. Second, the situation of an ectopic pelvic kidney has not been encountered by this study. Although the scanning time of the abdomen and pelvis is only 12 s, 2 kidneys which are markedly differing in their craniocaudal positioning may result in erroneous DRF data from this method. This method requires further prospective validation across the widest range of pathologies and patient body habitus between operators, centres and manufacturers before it can be applied generally.
View larger version (61K):
[in this window]
[in a new window]
|
Figure 6. (a) Contrast enhanced multidetector CT (CEMDCT) coronal reformat of patient 12 showing similar attenuation of enhancing renal tissue and enhancing renal cell carcinoma. The volume and attenuation data would have included the tumour. (b) Early 99Tcm-mercaptoacetyltriglycine (MAG3) renography images show exclusion of the renal cell carcinoma. The described method for CEMDCT derived differential renal function (DRF) will overestimate the contribution made by the RCC containing kidney.
|
|
|
Conclusion
|
|---|
It is possible to derive DRF from routine CEMDCT studies which are statistically equivalent to NM derived DRF across a wide range of clinical conditions and a wide range of DRF by the use of this simple post-processing technique. If this technique could be shown to be robust across all manufacturers and conditions, it may be possible for CEMDCT to routinely provide both anatomical and functional information for urological patients, avoiding the need for duplication with nuclear medicine investigations and saving both unnecessary radiation exposure to the patient and valuable imaging resources.
Received for publication January 10, 2006.
Revision received July 1, 2006.
Accepted for publication July 5, 2006.
|
References
|
|---|
- Dawson P, Peters AM. The use of contrast-enhanced computed tomography for the study of renal function and physiology. Invest Radiol 1993;28:S79–84.
- Miles KA, Leggett DAC, Bennett GAJ. CT derived Patlak images of the human kidney. Br J Radiol 1999;72:153–8.[Abstract]
- Tsushima Y, Blomley MJ, Kusano S, Endo K. Use of contrast-enhanced computed tomography to measure clearance per unit renal volume: a novel measurement of renal function and fractional vascular volume. Am J Kidney Dis 1999;33:754–60.[Medline]
- Tsushima Y. Functional CT of the kidney. Eur J Radiol 1999;30:191–7.[CrossRef][Medline]
- Frennby B, Almen T. Use of spiral CT and the contrast medium iohexol to determine in one session aortorenal morphology and the relative glomerular filtration rate of each kidney. Eur Radiol 2001;11:2270–7.[CrossRef][Medline]
- Nilsson H, Wadstrom J, Andersson LG, Raland H, Magnusson A. Measuring split renal function in renal donors: can computed tomography replace renography? Acta Radiologica 2004;4:474–80.[CrossRef]
- El-Diasty TA, Shokeir AA, Abo El-Ghar ME, Gad HM, Rafaie AF, Shehab El-Dein AB. Contrast enhanced spiral computerised tomography in live kidney donors: a single session for anatomical and functional assessment. J Urol 2004;171:31–4.[CrossRef][Medline]
- El-Ghar ME, Shokeir AA, El-Diasty TA, Refaie HF, Gad HM, El-Dein AB. Contrast enhanced spiral computerized tomography in patients with chronic obstructive uropathy and normal serum creatinine: a single session for anatomical and functional assessment. J Urol 2004;172:985–8.[CrossRef][Medline]
- Björkman H, Eklöf H, Wadström J, Andersson L-G, Nyman R, Magnusson A. Split renal function in patients with suspected renal artery stenosis: a comparison between gamma camera renography and two methods of measuremenet with computed tomography. Acta Radiol 2006;47:107–13.[CrossRef][Medline]
- Widjaja E, Oxtoby JW, Hale TL, Jones PW, Harden PN, McAll IW. Ultrasound measured renal length versus low dose CT volume in predicting single kidney glomerular filtration rate. Br J Radiol 2004;77:759–64.[Abstract/Free Full Text]
Top
Abstract
Introduction
