Professional Documents
Culture Documents
Hidronefrosis 6 PDF
Hidronefrosis 6 PDF
0022-5347/16/1954-1093/0 http://dx.doi.org/10.1016/j.juro.2015.10.173
THE JOURNAL OF UROLOGY
2016 by AMERICAN UROLOGICAL ASSOCIATION EDUCATION AND RESEARCH, INC.
Vol. 195, 1093-1099, April 2016
Printed in U.S.A.
www.jurology.com j 1093
1094 QUANTITATIVE ULTRASOUND FOR PEDIATRIC HYDRONEPHROSIS
techniques rely on analysis of digital images and and 18 females) with hydronephrosis who had undergone
creation of normative models of shape and appear- concurrent renal 2-D ultrasound and DR (MAG3 scan).
ance to which a test case is compared. The test case Mean patient age was 9.6 months (median 4.0, range 0 to
can then be defined, based on probability, as being 168). Of the cases 21 were left sided, 29 were right sided
and 8 were bilateral. Cases involving vesicoureteral
within a specific range of values different from the
reflux, ureterovesical junction obstruction, ureterocele
norm. By using machine learning to correlate pat-
or hydroureter/megaureter were excluded to avoid con-
terns of hydronephrosis with the diuretic renog- founding factors. Ultrasound acquisition was performed
raphy parameters, in this case t1/2 washout, the using the Philips iU22 (C2-9-D transducer) and GE
correlation between structure and function can be Logiq E9 (C8-5 transducer) scanners between July 2012
defined and used to categorize subsequent test and January 2013. Images were identified that contained
cases. With the primary goal being to identify ul- the closest to a full coronal view that included the renal
trasound derived thresholds of safety for hydro- pelvis and calyces. MAG3 scan was performed using a
nephrotic renal units where diuretic renography standardized method described previously.9,10 All pa-
can be avoided based on washout times, we applied tients underwent DR based on clinician preference, and
quantitative imaging techniques to a cohort of no study was postoperative.
pediatric patients with varying degrees of hydro-
nephrosis in whom routine ultrasonography and Quantitative Imaging Analysis
Manual segmentation of the renal parenchyma and col-
diuretic renography had been performed. The
lecting system was performed to create a cohort for pro-
quantitative imaging algorithm was then assessed
gram calibration and algorithm development. This was
as a predictor of the washout parameters of diuretic accomplished by manually drawing the best line around
renography. the perimeter of the renal parenchyma and the collecting
Currently the most widely used instrument for system (using the ITK-Snap Medical Image Segmentation
grading hydronephrosis sonographically is the SFU Tool, http://www.itk-snap.org), excluding the extrarenal
system,3 although it is not universally accepted as pelvis. The intrarenal pelvis was defined as the area
useful.4,5 Shapiro6 and Venkatesan7 et al proposed a falling within a line drawn around the hilar curve of the
quantitative measure of hydronephrosis severity, parenchyma (fig. 1). All segmentations were reconciled
the HI, defined as the ratio of the collecting system with 2 observers (pediatric radiologist and urologist) and
area to the total area of the collecting system submitted for QI analysis. The same cohort underwent
grading using the SFU classification by a pediatric radi-
and renal parenchyma. However, the practical use
ologist and a pediatric urologist.
of this index remains unclear.8 While different cli- A total of 131 morphological parameters were
nicians use differing thresholds for selecting pa- computed (fig. 2 and supplementary table, http://jurology.
tients in whom to perform DR and for defining com/). These descriptors can be intuitively grouped into
clinically significant obstruction, it would seem 3 categories, ie size, geometric shape and curvature. Size
valuable to perform DR only in those in whom it is descriptor parameters describe the size of the collecting
likely to reveal clinically relevant parameters for system and renal parenchyma, such as relative area
intervention or close followup. (eg HI) and perimeter, depth of calyces, and thickness of
We hypothesized that hydronephrosis can be parenchyma. Geometric shape descriptor parameters
objectively quantified sonographically using mod- include such factors as circularity ratio, eccentricity of the
ern imaging analysis, and that there is a correla- circumscribed ellipse and convexity. Curvature de-
scriptors are parameters that describe the local curvature
tion between degree of hydronephrosis and
of the structures under analysis (ie kidney and collecting
severity of obstruction as measured by DR. This
system).11
correlation likely is not linear, but demonstrates These quantitative shape features were used as po-
a threshold of dilatation above which clinically tential predictive variables to evaluate the diagnostic
significant drainage abnormalities are detectable performance of 2 different supervised classification
on DR. If this relationship can be defined, then models, the LOG12 and SVM with nonlinear kernel.13
quantitative assessment of degree of hydro- Both approaches have been widely used in the field of
nephrosis on US could be used to selectively iden- computer-aided diagnosis with satisfactory results.14
tify which patients have more severe obstruction These 2 classification approaches were applied to the
(by DR parameters) and should undergo DR, and test cohort to compute US based specific thresholds to
which patients have less severe hydronephrosis identify patients with a t1/2 above different thresholds,
ie 20, 30 and 40 minutes, since not all clinicians agree
and should forgo DR.
as to the clinically relevant DR thresholds. These cutoffs
represent commonly used clinical thresholds.15 The
design parameters of the model were adjusted to maxi-
MATERIALS AND METHODS mize the sensitivity of detecting severe cases of hydro-
Study Population nephrosis, that is no case with t1/2 above the threshold
Under an internal review board approved protocol we was misclassified. A best fit model was then derived for
retrospectively selected 50 consecutive patients (32 males each threshold using optimal morphological parameters
QUANTITATIVE ULTRASOUND FOR PEDIATRIC HYDRONEPHROSIS 1095
Figure 1. 2-D ultrasound slice selection and segmentation. A, selected slice containing whole longitudinal section of kidney.
B, consensus manual delineation of renal parenchyma and collecting system.
and SVM or LOG to categorize the renal units and classifiers, SVM and LOG (significance level 0.05).
severity of obstruction. Effect sizes were calculated using the Hedges formulation
To compare this QI method to the current SFU grading for means differences.19
system in selecting cases where DR was to be performed,
we compared SFU grade 3 and above or only SFU grade 4
as the criterion for performing DR. Sensitivity and spec- RESULTS
ificity were determined, as well as ROC curve analysis The retrospective power analysis of the cohort of
with AUC calculation. A similar assessment was made
50 patients provided a power value of greater than
using the previously described HI. The proposed frame-
0.99 in all instances. SFU grades ranged from 1 to 4
work was implemented in MATLAB, using a 64-bit
2.8 GHz processor, with an average execution time of (median 3). Median differential uptake on MAG3
12 seconds per case. scan was 48% (range 14% to 100%). Median t1/2 was
17 minutes (range 3 to greater than 120). Differen-
Statistical Analysis tial uptake was less than 40% in 7 patients and less
The cohort data were used to assess the performance than 30% in 3. All patients with differential uptake
of the QI algorithm described previously,2 using leave- less than 30% would have been identified with a t1/2
one-out cross-validation methodology.16 From the 131 threshold of 30 minutes or greater. Two patients
morphological features computed for each case an optimal with differential uptake less than 40% and greater
subset of predictive variables was selected for each t1/2 than 30% had washout times of less than 20 min-
threshold using the feature selection technique proposed utes (table 1), suggesting a nonobstructive cause of
by Cai et al.17 The significance of the improvement ob- the reduced uptake.
tained by the QI method over the SFU grade and HI was
The best performance of the QI method was ob-
statistically verified using the McNemar test.18 Addi-
tained by means of the SVM classification system
tionally the interobserver variance for SFU grading was
assessed using the Fleiss kappa statistic. for thresholds of 20 and 30 minutes, while LOG
In the absence of previous studies that related to this provided better performance for a threshold of
topic to provide an estimation of the expected effect 40 minutes. Because the design parameters of the
we performed a retrospective power analysis using a model were defined to maximize the sensitivity of
2-tailed t-test over the output probability provided by the detecting cases of hydronephrosis over a specified
Figure 2. Morphological descriptors of renal unit. Sample illustrates morphological descriptors considered as potential predictive
features of hydronephrosis severity. ab, eccentricity of circumscribed ellipse. Pmax, maximum parenchymal thickness. Pmin,
minimum parenchymal thickness. PKMAmax, kidney medial axis and point of maximum curvature. PKMAleft, left axis of kidney.
PKMAright, right axis of kidney. Arrows indicate curvature measures of collecting system and kidney surface. Relative area of
collecting system and parenchyma is shown.
1096 QUANTITATIVE ULTRASOUND FOR PEDIATRIC HYDRONEPHROSIS
Table 1. Cases by category that obtained using SFU grade when an SFU grade
No. Cases of 3 or 4 was used to refer patients for MAG3 im-
1
aging. For the 3 different t1/2 thresholds under
T /2 threshold:
20 Mins or greater 19
study the QI algorithm maximizes the reduction
30 Mins or greater 14 in MAG3 scan within the total cohort, while no
40 Mins or greater 8 critical case with severe obstruction was missed.
SFU grade:
2 14
3 18
4 18 DISCUSSION
Using the QI algorithm with machine learning in
pediatric patients with hydronephrosis, US can be
DR parameter, no case with t1/2 above the threshold used to predict which patients will have clinically
was misclassified (100% sensitivity). relevant increased washout times on DR. Applying
For the 3 thresholds considered, ie 20, 30 and this technique to children with hydronephrosis,
40 minutes, and at 100% sensitivity the specificities particularly those with prenatally diagnosed dis-
of the method were 94%, 70% and 74%, respectively. ease, could safely decrease the need for invasive and
Compared to using the SFU grading system to select ionizing DR. Clinicians would be able to select
patients for MAG3 scan (table 2) or the HI, this thresholds of safety that they find acceptable.
approach provides greater specificity while main- These results have several potentially useful
taining maximum sensitivity. implications. These observations demonstrate that
Figure 3 illustrates the ROC curves for the renal structure and function, specifically drainage,
quantitative imaging, SFU and hydronephrosis are related. Being able to use ultrasound images to
index techniques for selecting patients with t1/2 infer functional data would offer decreased inva-
greater than 30 minutes. Using either hydro- siveness, complexity and cost, with no reduction in
nephrosis index or SFU grading system showed an safety and acceptable reduction in accuracy.
AUC of 0.78, which was significantly less robust This system appears to be a useful clinical tool to
than the 0.94 AUC using the SVM based quantita- determine which patients would benefit from DR
tive imaging method (p <0.001 and p 0.04, over using the SFU scoring system alone. While this
respectively). Similar results were obtained for approach does not eliminate the need for DR, it
washout threshold values of 20 and 40 minutes. permits more specific application of a costly, com-
Quantitative imaging performed significantly better plex, semi-invasive imaging test with radiation
than hydronephrosis index (p <0.001 for either exposure. Application of this modality would be
threshold) and significantly better than the SFU most useful in infants with prenatally detected
grading system (p <0.001 for t1/2 of 20 minutes and hydronephrosis, in whom there remains controversy
p 0.05 for t1/2 of 40 minutes). The interobserver regarding the need for imaging beyond the post-
reliability for SFU grading between a pediatric natal US. There are currently no clearly defined
radiologist and a pediatric urologist revealed a guidelines based on objective metrics or clinical
Fleiss kappa statistic of 0.378, indicating fair outcomes that aid in the selection of patients for DR.
agreement.20 Of course, interpretation of the DR remains
Table 3 outlines the potential reduction in controversial, as is the decision regarding the need
MAG3 scanning using the QI method, compared to and timing for surgical intervention. This study
Table 2. QI vs SFU grade or HI to identify washout time thresholds in patients with hydronephrosis
% Sensitivity (95% CI) % Specificity (95% CI) % Pos Predictive Value (95% CI) % Accuracy (95% CI) % AUC (95% CI)
1
T /2 greater than 20 mins:
QI (SVM) 100 (82e100) 94 (79e99) 90 (70e99) 96 (86e99) 98 (92e100)
SFU grade 2 100 (82e100) 0 (0e11) 38 (25e53) 38 (25e53) 74 (57e83)
HI 100 (82e100) 52 (33e70) 56 (38e73) 70 (55e82) 77 (58e86)
T1/2 greater than 30 mins:
QI (SVM) 100 (77e100) 70 (52e84) 56 (35e75) 78 (64e88) 94 (83e98)
SFU grade 3 100 (77e100) 39 (23e56) 39 (23e56) 56 (41e70) 78 (62e86)
HI 100 (77e100) 47 (30e64) 42 (25e61) 62 (47e75) 78 (55e88)
T1/2 greater than 40 mins:
QI (LOG) 100 (63e100) 74 (58e86) 42 (20e66) 78 (64e88) 94 (81e98)
SFU grade 3 100 (63e100) 33 (20e50) 22 (10e39) 44 (30e59) 88 (56e90)
HI 100 (63e100) 62 (46e76) 33 (16e55) 68 (53e80) 80 (46e93)
For each case (t1/2 greater than 20, 30 or 40 minutes) SFU grade that resulted in 100% sensitivity was used. Optimized threshold of HI was used for prediction of t1/2 greater
than 20, 30 or 40 minutes. Accuracy equals true positive true negative/total positive total negative. Due to limited number of cases, exact 95% confidence interval is
provided (using binomial distribution).
QUANTITATIVE ULTRASOUND FOR PEDIATRIC HYDRONEPHROSIS 1097
REFERENCES
1. Siegel MJ: Pediatric Sonography, 4th ed. Phila- 11. Kovalevsky V: Curvature in digital 2D images. Int percentage differential function. J Pediatr Urol
delphia: Lippincott Williams & Wilkins 2010. J Patt Recogn Artif Intell 2001; 15: 1183. 2007; 3: 382.
2. Cerrolaza JJ, Peters CA, Martin AD et al: 12. Hosmer DW and Lemeshow S: Applied Logistic 22. Dias CS, Silva JM, Pereira AK et al: Diagnostic
Ultrasound based computer-aided-diagnosis of Regression, 2nd ed. New York: John Wiley & accuracy of renal pelvic dilatation for detecting
kidneys for pediatric hydronephrosis. Proc SPIE Sons Inc. 2000. surgically managed ureteropelvic junction
9035, Medical Imaging 2014; 90352T-1. obstruction. J Urol 2013; 190: 661.
13. Cristianini N and Shawe-Taylor J: An Introduc-
3. Fernbach SK, Maizels M and Conway JJ: Ultra- tion to Support Vector Machines and Other
sound grading of hydronephrosis: introduction to 23. Cost GA, Merguerian PA, Cheerasam SP et al:
Kernel-Based Learning Methods. New York: Sonographic renal parenchymal and pelvicaliceal
the system used by the Society for Fetal Urology. Cambridge University Press 2000.
Pediatr Radiol 1993; 23: 478. areas: new quantitative parameters for renal
14. Sun T, Wang J, Li X et al: Comparative evalua- sonographic followup. J Urol 1996; 156: 725.
4. Keays MA, Guerra LA, Mihill J et al: Reliability
tion of support vector machines for computer
assessment of Society for Fetal Urology ultra- 24. Rodrguez LV, Lock J, Kennedy WA et al: Evalu-
aided diagnosis of lung cancer in CT based on a
sound grading system for hydronephrosis. J Urol, ation of sonographic renal parenchymal area in
multi-dimensional data set. Comput Methods
suppl., 2008; 180: 1680. the management of hydronephrosis. J Urol 2001;
Programs Biomed 2013; 111: 519.
165: 548.
5. Kim SY, Kim MJ, Yoon CS et al: Comparison of
15. Eskild-Jensen A, Gordon I, Piepsz A et al:
the reliability of two hydronephrosis grading 25. Pulido JE, Furth SL, Zderic SA et al: Renal
Interpretation of the renogram: problems and
systems: the Society for Foetal Urology grading parenchymal area and risk of ESRD in boys with
pitfalls in hydronephrosis in children. BJU Int
system vs. the Onen grading system. Clin Radiol posterior urethral valves. Clin J Am Soc Nephrol
2004; 94: 887.
2013; 68: e484. 2014; 9: 499.
6. Shapiro SR, Wahl EF, Silberstein MJ et al: 16. Elisseeff A and Pontil M: Leave-one-out error
Hydronephrosis index: a new method to track and stability of learning algorithms with appli- 26. Riccabona M, Fritz GA, Schollnast H et al: Hydro-
patients with hydronephrosis quantitatively. cations. In: Advances in Learning Theory: nephrotic kidney: pediatric three-dimensional
Urology 2008; 72: 536. Methods, Models and Applications. Edited by US for relative renal size assessmentdinitial
J Suykens, G Horvath, S Basu et al: Amsterdam: experience. Radiology 2005; 236: 276.
7. Venkatesan KJ, Green J, Shapiro SR et al: IOS Press 2003; chapt 6, pp 111e129.
Correlation of hydronephrosis index to Society 27. Kang X, Safdar N, Myers E et al: Automatic
of Fetal Urology hydronephrosis scale. Adv Urol 17. Cai D, Zhang C and He X: Unsupervised feature analysis of pediatric renal ultrasound using
2009; 960490. selection for multi-cluster data. Presented at shape, anatomical and image acquisition priors.
annual conference of ACM SIGKDD, Washington, Med Image Comput Comput Assist Interv 2013;
8. Rud O, Moersier J, Peter J et al: Prospective D.C., July 25-28, 2010. 8151: 259.
evaluation of interobserver variability of the
hydronephrosis index and the renal resisitive 18. Bostanci B and Bostanci E: An evaluation of 28. Cerrolaza JJ, Safdar N, Peters CA et al: Seg-
index as sonographic examination methods for classification algorithms using McNemars test. mentation of kidney in 3D-ultrasound images
the evaluation of acute hydronephrosis. BJU Int Presented at International Conference on Bio- using Gabor-based appearance models. Presented
2012; 110: E350. Inspired Computing: Theories and Applications, at IEEE International Symposium on Biomedical
Gwalior, India, December 14-16, 2012. Imaging, Beijing, China, April 29-May 2, 2014.
9. Shulkin BL, Mandell GA, Cooper JA et al:
Procedure guideline for diuretic renography in 19. Hedges LV: Distribution theory for Glasss esti-
children 3.0. J Nucl Med Technol 2008; 36: 162. 29. Hafez AT, McLorie G, Bagli D et al: Analysis of
mator of effect size and related estimators.
trends on serial ultrasound for high grade
J Educ Behav Stat 1981; 6: 107.
10. Conway JJ and Maizels M: The well neonatal hydronephrosis. J Urol 2002; 168: 1518.
tempered diuretic renogram: a standard method 20. Landis JR and Koch GG: The measurement of
to examine the asymptomatic neonate with observer agreement for categorical data. Bio- 30. Ross SS, Kardos S, Krill A et al: Observation of
hydronephrosis or hydroureteronephrosis. A metrics 1977; 33: 159. infants with SFU grades 3-4 hydronephrosis:
report from combined meetings of the Society worsening drainage with serial diuresis renog-
for Fetal Urology and members of the Pediatric 21. Erickson BA, Maizels M, Shore RM et al: raphy indicates surgical intervention and helps
Nuclear Medicine Councildthe Society of Newborn Society of Fetal Urology grade 3 prevent loss of renal function. J Pediatr Urol
Nuclear Medicine. J Nucl Med 1992; 33: 2047. hydronephrosis is equivalent to preserved 2011; 7: 266.
EDITORIAL COMMENT
on ultrasonographic images, may lead to more of renography. Future studies will hopefully give us
objective and, perhaps most importantly, reproduc- the potential to assess differential renal function
ible results of diagnosis and outcome measures. using 3-D quantitative imaging as a noninvasive
So far, semiquantitative US has been related to tool without the exposure to radiation.
t1/2 on MAG3 renography, a quite controversial
L. Henning Olsen
parameter, especially in younger children and in- Department of Urology, Section of Pediatric Urology
fants. To some extent, this method is observer Aarhus University Hospital
dependent but the same is the case in assessment Aarhus, Denmark