ESPU Meeting on Thursday 20, April 2023, 10:50 - 11:45
10:50 - 10:53
S10-1 (OP)
Sang Hoon SONG 1, Kun Suk KIM 1, Sang Don LEE 2, Jae Min CHUNG 2, Young In KIM 3, Hyejung YOUN 4 and Jihoon KWEON 4
1) Asan Medical Center, University of Ulsan College of Medicine, Department of Urology, Seoul, REPUBLIC OF KOREA - 2) Pusan National University Yangsan Hospital, Pusan National University School of Medicine, Department of Urology, Yangsan, REPUBLIC OF KOREA - 3) Asan Medical Institute of Convergence Science and Technology, Asan Medical Center, Department of Medical Science, Seoul, REPUBLIC OF KOREA - 4) Asan Medical Center, Department of Convergence Medicine, Seoul, REPUBLIC OF KOREA
PURPOSE
We aimed to externally validate previously developed and trained algorithms to predict kidney area and hydronephrosis area from ultrasonography images in pediatric patients.
MATERIAL AND METHODS
A cohort was composed of 60 pediatric patients with 112 longitudinal ultrasound images with or without hydronephrosis from a Hospital, which is independent of the institution where the algorithms were developed. In this cohort, we analyzed the accuracy of the previously developed algorithms for their segmentation performance using dice similarity coefficients (DSC) by comparison with the manually labeled area.
RESULTS
The mean anteroposterior pelvic diameter was 4.8 (0-40) mm. SFU grade was 0 in 21 (18.8%), Ⅰ in 41 (36.6%), Ⅱ in 35 (31.3%), Ⅲ in 9 (8.0%), and Ⅳ in 6 (5.4%). The DSC of the 10 models ensemble was 0.90 on average, 0.84 for hydronephrosis area, and 0.95 for kidney outline. After excluding images without hydronephrosis, the DSC of the 10 models ensemble was 0.90 on average, 0.84 for the hydronephrosis area, and 0.95 for the kidney outline. When we analyzed the images with SFU gr 3 and 4, the DSC of the 10 models ensemble was 0.93 on average, 0.91 for the hydronephrosis area, and 0.96 for the kidney outline. The difference in DSC from the development cohort was only -0.01 on average, -0.02 for the hydronephrosis area, and -0.01 for the kidney outline.
CONCLUSIONS
Our algorithm showed high performance in segmenting kidney and hydronephrosis areas on external datasets, demonstrating its generalizability to diverse ultrasonography images especially with moderate to severe hydronephrosis.
10:53 - 10:56
S10-2 (OP)
Adree KHONDKER 1, Jethro KWONG 2, Jin Kyu KIM 3, Michael CHUA 3, Margarita CHANCY 3, Natasha BROWNRIGG 3, Joana DOS SANTOS 1, Lauren ERDMAN 4, Neeta D'SOUZA 5, John WEAVER 6, Mandy RICKARD 3 and Armando LORENZO 3
1) The Hospital for Sick Children, Division of Urology, Toronto, CANADA - 2) University of Toronto, Division of Urology, Toronto, CANADA - 3) The Hospital for Sick Children, Urology, Toronto, CANADA - 4) The Hospital for Sick Children, Computer Science, Toronto, CANADA - 5) SUNY Downstate, School of Medicine, New York, USA - 6) UH Rainbows Cleveland, Division of Urology, Cleveland, USA
PURPOSE
To use routinely reported ultrasound findings, along with machine learning approaches, to predict the risk of ureteropelvic junction obstruction (UPJO) on diuretic renogram in infants with isolated HN.
MATERIAL AND METHODS
We included patients less than 24 months of age at baseline with a renogram within 3 months of an ultrasound. Age, sex, and routinely reported ultrasound findings (laterality, kidney length, anteroposterior diameter [APD], SFU grade) were abstracted. T 1/2 washout periods were collected from renography and stratified as low risk (<20 minutes), unclear risk (20-60 minutes), and high risk of obstruction (>60 minutes). A random forest model was trained to classify obstruction risk, referred to as AI Evaluation of Renogram Obstruction (AERO). Model performance was determined by measuring area under the receiver-operator-characteristic curve (AUROC) and decision curve analysis.
RESULTS
A total of 304 patients met inclusion criteria, with a median age of diuretic renogram at 4 months (IQR 2, 7). Of all patients, 48 (16%) were low-risk, 102 (33%) were unclear risk, 156 (51%) were high risk of obstruction based on diuretic renogram. AERO achieved a multi-class AUROC of 0.75 which was superior to logistic regression. The most important features for prediction included age, APD, and SFU grade. We deployed our model in an easy-to-use application (https://sickkidsurology.shinyapps.io/AERO/). At a threshold cutoff of 30%, AERO would allow 137 more patients per 1000 to safely avoid a renogram without missing significant obstruction compared to a strategy in which a renogram is performed for all patients with SFU grade ≥ 3.
CONCLUSIONS
Routine ultrasound findings can be used to determine if a diuretic renogram can be safely avoided for children with isolated hydronephrosis, thus offering the potential to minimize invasiveness of monitoring and exposure to radiation.
10:56 - 10:59
S10-3 (OP)
C. D. Anthony HERNDON 1, Sarah WILLIAMSON 2, Carol DAVIS-DAO 3, Melissa MCGRATH 4, Antoine KHOURY 3, Gina LOCKWOOD 5, Rebecca ZEE 1, Nora KERN 6, Valre WELCH 1, Anne DUDLEY 7, Shannon CANNON 8, Kai-Wen CHUANG 3, Heidi STEPHANY 3, Elias WEHBI 3 and Luis BRAGA 4
1) Children's Hospital of Richmond at Virginia Commonwealth University, Department of Urology, Richmond, USA - 2) Children's Hospital of the King's Daughters, Division of Urology, Norfolk, USA - 3) Children's Hospital of Orange County; University of California at Irvine, Division of Pediatric Urology, Orange, USA - 4) McMaster University, Department of Urology, Hamilton, CANADA - 5) University of Iowa, Department of Urology, Iowa City, USA - 6) University of Virginia, Department of Urology, Charlottesville, USA - 7) Connecticut Children's Medical Center, Division of Urology, Hartford, USA - 8) University of Wisconsin, Department of Urology, Madison, USA
PURPOSE
Prenatal hydronephrosis is a common prenatal ultrasound finding and the majority of patients will resolve. Our objective was to evaluate the Hydronephrosis Outcome Prediction Score (HOPS), utilizing three ultrasound measurements to predict resolution.
MATERIAL AND METHODS
Patients with isolated hydronephrosis were prospectively enrolled from 7 centers. Baseline ultrasounds were included if performed 3 days to 6 months of life. HOPS (adapted from Li et al. Front Pediatr 2020;8:35) utilizes three ultrasound parameters: SFU grade, anterior posterior diameter (APD), and percentage difference of renal lengths. HOPS was calculated on 12-point scale, each parameter assigned score 0 (least severe) to 4 (most severe). Hydronephrosis resolution defined as SFU grade 0/1 and APD ≤10 mm. Area Under the Curve (AUC) and Receiver Operating Characteristic Curves were determined to establish HOPS thresholds for resolution.
RESULTS
Out of 1983 patients enrolled, 601 met inclusion criteria. Median age was 1.4 months, and median follow-up was 23 months. 49% resolved, 19% underwent pyeloplasty, and 32% continued follow-up. Median HOPS for resolution was 3.0, 5.0 for continued follow-up, and 9.5 for pyeloplasty. HOPS of ≤4 was a threshold for resolution (AUC=0.75). Of patients with HOPS ≤4, 74% resolved, only 2% underwent pyeloplasty; of the 24% that remained under follow-up, none had SFU grade 4 at most recent ultrasound.
CONCLUSIONS
HOPS of ≤4 on initial ultrasound predicted which patients will have spontaneous resolution of isolated hydronephrosis. Our findings suggest that patients with HOPS ≤4 are at low risk for progression as only 2% underwent surgery for obstruction. HOPS is easily measured from ultrasound data and can be used to tailor diagnostic studies and clinical follow-up.
10:59 - 11:02
S10-4 (OP)
Lauren ERDMAN 1, Stanley Bryan HUA 2, Kunj SHETH 3, Daniel ALVAREZ 3, Mandy RICKARD 4, Armando LORENZO 4, Kyla N VELEAR 5, Anna GOLDENBERG 6, Joana DOS SANTOS 7 and Michael CHUA 8
1) Hospital for Sick Children, University of Toronto, Cincinnati Children's Hospital, Center for Computational Medicine, Computer Science, Anderson Center, and Gastroenterology, Toronto, CANADA - 2) Hospital for Sick Children, University of Toronto, Center for Computational Medicine, Computer Science, Toronto, CANADA - 3) Stanford Children's Health, Urology, Sunnyvale, USA - 4) Hospital for Sick Children, University of Toronto, Surgery, Toronto, CANADA - 5) Stanford Children's Health, Surgery, Palo Alto, USA - 6) Hospital for Sick Children, University of Toronto, Computer Science, Toronto, CANADA - 7) Hospital for Sick Children, University of Toronto, Urology, Toronto, CANADA - 8) Hospital for Sick Children, University of Toronto, Urology, Surgery, Toronto, CANADA
PURPOSE
Accurate, early prediction of obstructive hydronephrosis (HN) could dramatically reduce the number of clinic visits and additional investigations for non-obstructed patients, which comprises approximately 80% of this population. In this work we assessed a previously developed artificial intelligence (AI) model which predicts obstructive HN directly from a 1 sagittal and 1 transverse renal ultrasound image. Herein, we ask if this prediction improves with time (i.e. as the patient ages) or with serial ultrasounds over time.
MATERIAL AND METHODS
Our AI model is developed using retrospectively collected data from 1938 ultrasound images for 403 patients and their linked health record. We used methods from video action recognition to adapt our original single-visit to a multi-visit model and compare the performance of these models using area under the receiver operator curve (AUROC) and area under the precision-recall curve (AUPRC). We test this model both in a test Source data set and in data from a second institution in North America, Institution 2.
RESULTS
We find >90% AUROC in all our models. In our Source test set, the AUROC using convolutional pooling (0.97) and LSTM (0.97) is significantly higher than the AUROC using the first ultrasound (0.93) only or averaging all predictions (0.93). However, the high AUROC (0.93) even using the first ultrasound in our Source test data indicates the utility of this model even very early on.
CONCLUSIONS
The benefit of additional ultrasounds in our model is significant but small. Early decision-making with AI tools such as the one presented here can streamline the care of very-low-risk patients, while additional follow-up visits can be obtained for high-risk or more complex cases.
11:17 - 11:20
S10-5 (OP)
Saidanvar AGZAMKHODJAEV 1, Zafar ABDULLAEV 2, Akmal RAKHMATULLAEV 2 and Sergey BONDARENKO 3
1) National Children's Medical Center, Paediatric Urology, Tashkent, UZBEKISTAN - 2) National Children's Medical Center, Pediatric urology, Tashkent, UZBEKISTAN - 3) 7th Municipal Hospital, Pediatric urology, Volgograd, RUSSIAN FEDERATION
PURPOSE
To asses the ultrasound parameters to identify the reliable diagnostic parameters for predicting the postnatal surgery in fetuses with isolated hydronephrosis.
MATERIAL AND METHODS
From 2015 to 2019, a prospective analysis of 60 fetuses (62 renal units) with isolated grade III hydronephrosis was conducted. The study included the measurement of the diameter of the calyces, the anteroposterior diameter of the renal pelvis (APDRP), and the shape of the pelvis (funnel-shaped or elliptical) was recorded. With an elliptical pelvis, the compression ratio of the ellipse was calculated as the ratio of the semi-minor axis of the ellipse to its semi-major axis. All parameters were recorded in the third trimester of pregnancy. The study did not include megaureter, VUR, ureterocele, duplication of the kidney and posterior urethral valve.
RESULTS
When analyzing the ROC-curves, it was found that the APDRP has a low quality of the binary classification, the area under the curve was 0.62 (p = 0.12), while the shape of the pelvis and the compression ratio in the elliptical shape turned out to be better diagnostic tests (the area under the curve is 0.79 and 0.93, p = 0.000 and 0.000, respectively). In patients with an elliptical pelvis, the chance of aggravation of the grade of hydronephrosis and APDRP in the postnatal period increased with an increase in the compression ratio (test sensitivity 92.3%, specificity 92.3%, p = 0.000). When analyzing the logistic curve, it was found that at the cut-off point of the compression ratio equal to or greater than 0.7, the probability of early surgery exceeded 80% (p = 0.006).
CONCLUSIONS
In patients with antenatal hydronephrosis an ultrasound parameters based on the shape of the renal pelvis have a high predictive value for predicting the postnatal surgery. This is important in selecting children for early surgery before the loss of renal function is irreversible.
11:20 - 11:23
S10-6 (OP)
Sibel TIRYAKI 1, Melis CEVHERTAS 2, Dilnur SEVINC 3, Omer Baris YUCEL 2, Bade TOKER KURTMEN 3, Ali TEKIN 2 and Ibrahim ULMAN 2
1) Tepecik Research and Training Hospital, Division of Pediatric Urology, Izmir, TÜRKIYE - 2) Ege University, Department of Pediatric Surgery, Division of Pediatric Urology, Izmir, TÜRKIYE - 3) Tepecik Training and Research Hospital, Pediatric Surgery, Izmir, TÜRKIYE
PURPOSE
Main indications for surgery during follow-up of UPJ-type congenital hydronephrosis are increase in hydronephrosis and loss of function. Despite the subjective nature of ultrasonographic evaluation, kidney length is regarded as a relatively more objective parameter. We aimed to determine whether monitoring the patient with a kidney length nomogram will provide data that helps decision-making for UPJ obstruction.
MATERIAL AND METHODS
Patients with sufficient data and without concomitant urologic abnormalities among the patients who underwent surgery for UPJ obstruction between 2012 and 2022 were included. Long axis diameter of the kidney in repetitive studies in preoperative follow-up were noted on an age and sex approparitate nomogram curve (Obrycki et al, 2022). Deviation from one's own percentile was regarded as a sudden increase in length. Concomitant sonographic findings (pelvis AP diameter, parenchymal thickness, hydronephrosis grade) were evaluated.
RESULTS
A total of 143 patients among 256 were included. A sudden increase in length was detected in preoperative sonography in 59 of these (F:19, M:40). The median age at operation was 11 (1-179) months. Surgery was decided at the time of the sonography showing a sudden increase in diameter in 56% of the patients. The remaining 26 patients were operated 6.5 (2-39) months after this time point. There was no significant change in either anteroposterior diameter (18.5 (7-46) vs. 20 (10-37); p=0.679) or parenchymal thickness (depicted as ratio with contralateral side, 0.73±0.27 vs. 0.74±0.26; p=0.894) at the sonography revealing sudden increase in length.
CONCLUSIONS
In our study, a sudden increase in kidney length was observed in a median of six months before the decision of surgery was made in a considerable amount of the patients. No significant change was observed in other sonographic findings evaluated during this period. Significant increase in kidney length during follow-up may indicate surgical treatment despite a stable pelvic AP diameter, parenchymal thickness and hydronephrosis grade.
11:23 - 11:26
S10-7 (OP)
Roseanne FERREIRA 1, Yaqoub JAFAR 1, Melissa MCGRATH 1, Carol DAVIS-DAO 2, Sarah HOLZMAN 2, Anthoine KHOURY 3, Anthony HERNDON 4 and Luis BRAGA 1
1) Mcmaster children hospital, Pediatric urology, Hamilton, CANADA - 2) UNIVERSITY OF IRVINE CALIFORNIA, Pediatric urology, Hamilton, CANADA - 3) UNIVERSITY OF IRVINE CALIFORNIA, Hamilton, USA - 4) University of Virginia, Pediatric urology, Hamilton, USA
PURPOSE
To determine the utility of the hydronephrosis outcomes prediction (HOP) score in identifying severe UPJO-like hydronephrosis suggestive of obstruction.
MATERIAL AND METHODS
A prenatal HN database(2008-22) was reviewed to select patients with UPJO-like HN. Those with VUR, megaureter and other anomalies were excluded. HOP score parameters (SFU grade, APD,absolute percentage diffrence of renal lengths at the first ultrasound), MAG3 t½ drainage time and curve patterns, and surgery were collected. Obstruction on a renogram was defined as a t½>30min and/or a non-descending (flat or ascending) drainage curve. Logistic regression, t-test or Wilcoxon-sum were used for statistical analysis.
RESULTS
Of the 421 UPJO-like HN patients with MAG3 selected, 323(77%) were male and 87(21%) had pyeloplasty. The mean age at baseline was 2.9 months(IQR 1.4-7.6). A HOP score of 8 was an optimal cut-off point for obstruction with 71% of sensitivity and 89% of specificity (AUC=0.897, 95%CI:0.86-0.93). Obstruction on renogram was strongly associated with surgery (66% vs. 1%p<0.001). Median HOP score in those who had a pyeloplasty was 10 vs. 5 in those managed non-surgically(p<0.01).
| Groups | N | Mean HOP score | P-value |
|
T1/2 (total) <=30 min >30 min |
420 359 61 |
6.1 5.5 9.4 |
<0.01 |
|
renal scan curve (total) -desending -asending or flat |
421 229 122 |
6.1 5 8.9 |
<0.01 |
|
Obstruction on renal scan (total) -non-obstructive ->30 min or non-descending |
421 294 127 |
6.1 5 8.8 |
<0.01 |
CONCLUSIONS
If HOP score >8(A) is associated with obstruction on renogram(B) and obstruction(B) is more likely to lead to surgical intervention(C), then one can logically conclude that HOP score >8(A) is also more likely to be associated with surgery(C), following the mathematical principle of A=B and B=C, therefore A=C. The HOP score accurately identified patients with severe HN, more likely to be obstructed, and, therefore, undergo surgery based on the aforementioned renogram criteria.
11:26 - 11:29
S10-8 (OP)
Ketaki GHARPURE, Sara LOBO, Navroop JOHAL and Abraham CHERIAN
Great Ormond Street Hospital, Urology, London, UNITED KINGDOM
PURPOSE
Renal pelvic diameter in the transverse plane is commonly used to quantify hydronephrosis. Inaccuracies arise due to interobserver variability, hydration status and pure intra-renal dilatation. We explored usefulness of "Differential renal length index" (iDRL) before and after pyeloplasty.
MATERIAL AND METHODS
Prospectively collected data of all children undergoing pyeloplasty for isolated unilateral PUJ hold up from 2015-2018 were analysed. Structural anomalies of the upper and lower urinary tract, bilateral involvement, and subnormal function (<20%) excluded. All had pre and postoperative Ultrasound scan and Tc99mMAG3 renograms. iDRL was calculated as below:
iDRL = [(A- B) ÷ B] x 100
Where A - Length of hydronephrotic kidney (cm)
B - Length of contralateral normal kidney (cm)
Data analysed by SPSS-v29. The mean difference and standard error of mean (SEM) between pre and post-operative iDRL was evaluated by paired Student's T-test, p<0.05 - significant.
RESULTS
A total of 53 renal units were included. Median age 8-months (Range 1-181). Males 64%, Laparoscopic 43%, Symptomatic PUJO 36%. Follow up was 1 year(n=53) and 3 years(n=38).
Preoperatively, mean iDRL: 28.5 (+/- 2.10 SEM).
Postoperatively, at 1-year follow-up, mean iDRL decreased to 9.9 (+/- 1.5 SEM). Mean improvement in iDRL: 18.6 (+/- 2.1 SEM)(p<0.001).
Postoperatively at 3-year follow-up, mean iDRL was 11.8 (+/-1.8 SEM). Mean improvement in iDRL: 20 (+/-2.6 SEM)( p<0.001).
All improvement was supported with imaging and symptom resolution.
CONCLUSIONS
Our pilot study identifies iDRL as a useful measure of improvement following successful pyeloplasty. Whether iDRL can be used in the preoperative assessment remains unanswered but paves the way to our next study.