Deep Learning of Videourodynamics to Classify Bladder Dysfunction Severity in Patients With Spina Bifida.

2023

994-1003

05/2023

1527-3792

PURPOSE: Urologists rely heavily on videourodynamics to identify patients with neurogenic bladders who are at risk of upper tract injury, but their interpretation has high interobserver variability. Our objective was to develop deep learning models of videourodynamics studies to categorize severity of bladder dysfunction.

MATERIALS AND METHODS: We performed a cross-sectional study of patients aged 2 months to 28 years with spina bifida who underwent videourodynamics at a single institution between 2019 and 2021. The outcome was degree of bladder dysfunction, defined as none/mild, moderate, and severe, defined by a panel of 5 expert reviewers. Reviewers considered factors that increase the risk of upper tract injury, such as poor compliance, elevated detrusor leak point pressure, and detrusor sphincter dyssynergia, in determining bladder dysfunction severity. We built 4 models to predict severity of bladder dysfunction: (1) a random forest clinical model using prospectively collected clinical data from videourodynamics studies, (2) a deep learning convolutional neural network of raw data from the volume-pressure recordings, (3) a deep learning imaging model of fluoroscopic images, (4) an ensemble model averaging the risk probabilities of the volume-pressure and fluoroscopic models.

RESULTS: Among 306 videourodynamics studies, the accuracy and weighted kappa of the ensemble model classification of bladder dysfunction when at least 75% expected bladder capacity was reached were 70% (95% CI 66%,76%) and 0.54 (moderate agreement), respectively. The performance of the clinical model built from data extracted by pediatric urologists was the poorest with an accuracy of 61% (55%, 66%) and a weighted kappa of 0.37.

CONCLUSIONS: Our models built from urodynamic pressure-volume tracings and fluoroscopic images were able to automatically classify bladder dysfunction with moderately high accuracy.

10.1097/JU.0000000000003267

J Urol

36787376

Caudal anesthesia is not associated with post-operative complications following distal hypospadias repair.

2023

03/2023

1873-4898

INTRODUCTION: The use of caudal anesthesia at the time of hypospadias repair remains controversial as some prior studies have reported increased postoperative complication rates associated with caudal nerve block. However, these results have been called into question due to confounding factors and heterogeneous study groups.

OBJECTIVE: Given the importance of identifying true risk factors associated with increased hypospadias complication rate, we examined our experience with caudal anesthesia limiting our analysis to distal repairs. We hypothesized that caudal anesthesia would not be associated with increased postoperative complications.

STUDY DESIGN: We retrospectively reviewed our institutional hypospadias database from June 2007 to January 2021. All boys who underwent single-stage distal hypospadias repair with either caudal or penile block with minimum 1 month follow up were included. Records were reviewed to determine the type of local anesthesia, type of hypospadias repair, all complications, and time to complication. Association between any complication and local anesthesia type was evaluated by univariate and multivariate logistic regression analysis controlling for age at surgery and type of repair. A sub-analysis was performed for complications occurring ≤30 days.

RESULTS: Overall, 1008 boys, 832 (82.5%) who received caudal and 176 (17.5%) penile block, were included. Median age at surgery was 8.1 months and median follow up was 13 months. Overall complication rate was 16.4% with 13.8% of patients requiring repeat operation. Median time to complication was 10.59 months and was significantly shorter in the caudal group (8.45 vs. 25.2 months). Caudal anesthesia was associated with higher likelihood of complication on univariate analysis; however, this was not true on multivariate analysis when controlling for age and type of repair. Caudal anesthesia was not associated with increased likelihood of complication within 30 days.

DISCUSSION: Since the association between caudal anesthesia and hypospadias complications was first suggested, several studies have tried to answer this question with variable results. Our findings add to the evidence that there is no association between caudal anesthesia and increased hypospadias complications in either the short or long term. The major strengths of our study are a large, homogenous study population, robust follow up and inclusion of data from 14 surgeons over 14 years. Limitations include the study's retrospective nature as well as lack of standardized follow up protocol throughout the study period.

CONCLUSIONS: After controlling for possible confounders, caudal nerve block was not associated with increased risk of postoperative complications following distal hypospadias repair.

10.1016/j.jpurol.2023.03.019

J Pediatr Urol

37002025

Deep Learning of Videourodynamics to Classify Bladder Dysfunction Severity in Patients With Spina Bifida.

2023

994-1003

05/2023

1527-3792

PURPOSE: Urologists rely heavily on videourodynamics to identify patients with neurogenic bladders who are at risk of upper tract injury, but their interpretation has high interobserver variability. Our objective was to develop deep learning models of videourodynamics studies to categorize severity of bladder dysfunction.

MATERIALS AND METHODS: We performed a cross-sectional study of patients aged 2 months to 28 years with spina bifida who underwent videourodynamics at a single institution between 2019 and 2021. The outcome was degree of bladder dysfunction, defined as none/mild, moderate, and severe, defined by a panel of 5 expert reviewers. Reviewers considered factors that increase the risk of upper tract injury, such as poor compliance, elevated detrusor leak point pressure, and detrusor sphincter dyssynergia, in determining bladder dysfunction severity. We built 4 models to predict severity of bladder dysfunction: (1) a random forest clinical model using prospectively collected clinical data from videourodynamics studies, (2) a deep learning convolutional neural network of raw data from the volume-pressure recordings, (3) a deep learning imaging model of fluoroscopic images, (4) an ensemble model averaging the risk probabilities of the volume-pressure and fluoroscopic models.

RESULTS: Among 306 videourodynamics studies, the accuracy and weighted kappa of the ensemble model classification of bladder dysfunction when at least 75% expected bladder capacity was reached were 70% (95% CI 66%,76%) and 0.54 (moderate agreement), respectively. The performance of the clinical model built from data extracted by pediatric urologists was the poorest with an accuracy of 61% (55%, 66%) and a weighted kappa of 0.37.

CONCLUSIONS: Our models built from urodynamic pressure-volume tracings and fluoroscopic images were able to automatically classify bladder dysfunction with moderately high accuracy.

10.1097/JU.0000000000003267

J Urol

36787376

Reply By Authors.

2023

1003

05/2023

1527-3792

10.1097/JU.0000000000003267.03

J Urol

37026634

Deep Learning of Videourodynamics to Classify Bladder Dysfunction Severity in Patients with Spina Bifida.

2023

101097JU0000000000003267

02/2023

1527-3792

PURPOSE: Urologists rely heavily on videourodynamics (VUDS) to identify patients with neurogenic bladders who are at risk of upper tract injury, but their interpretation has high interobserver variability. Our objective was to develop deep learning models of VUDS studies to categorize severity of bladder dysfunction.

MATERIALS AND METHODS: We performed a cross-sectional study of patients aged 2 months to 28 years with spina bifida who underwent VUDS at a single institution between 2019 and 2021. The outcome was degree of bladder dysfunction, defined as none/mild, moderate, and severe, defined by a panel of 5 expert reviewers. Reviewers considered factors that increase the risk of upper tract injury, such as poor compliance, elevated detrusor leak point pressure, and detrusor sphincter dyssynergia, in determining bladder dysfunction severity. We built 4 models to predict severity of bladder dysfunction: 1) a random forest clinical model using prospectively collected clinical data from VUDS studies; 2) deep learning convolutional neural network of raw data from the volume-pressure recordings; 3) deep learning imaging model of fluoroscopic images; 4) ensemble model averaging the risk probabilities of the volume-pressure and fluoroscopic models.

RESULTS: Among 306 VUDS studies, the accuracy and weighted kappa of the ensemble model classification of bladder dysfunction when at least 75% expected bladder capacity was reached were 70% (95% CI 66%,76%) and 0.54 (moderate agreement), respectively. The performance of the clinical model built from data extracted by pediatric urologists was the poorest with an accuracy of 61% (55%, 66%) and a weighted kappa of 0.37.

CONCLUSIONS: Our models built from urodynamic pressure-volume tracings and fluoroscopic images were able to automatically classify bladder dysfunction with moderately high accuracy.

10.1097/JU.0000000000003267

J Urol

36787376

Does ketorolac administration at the time of hypospadias surgery increase unplanned encounters in the immediate postoperative period?

2023

01/2023

1873-4898

INTRODUCTION & OBJECTIVE: The opioid crisis has raised concerns for long-term sequela of routine administration of opioids to patients, particularly in the pediatric population. Nonsteroidal anti-inflammatory drug use is limited in hypospadias surgery due to concerns for post-operative bleeding, particularly with ketorolac. We hypothesize that ketorolac administration at the time of hypospadias repair is not associated with increased bleeding or immediate adverse events.

METHODS: A retrospective single institution study included all patients undergoing hypospadias surgery from 2018 to 2021. Outcomes measured include peri-operative ketorolac administration, opioid prescriptions, and unplanned encounters (i.e., emergency department or office visits). Comparative statistics using non-parametric and binary/categorical tests and a logistic regression were performed.

RESULTS: 1044 patients were included, among whom there were 562 distal, 278 proximal and 204 hypospadias complication repairs. Ketorolac was administered to 396 (37.9%) patients and its utilization increased during the study period [Summary Figure]. Patients receiving ketorolac were older (p = 0.002) and were prescribed opioids less often after surgery (2.0% vs 5.2%, p = 0.009). There was no difference in unplanned encounters across repair types (p = 0.1). Multivariate logistic regression showed ketorolac use was not associated with an increased likelihood of an unplanned encounter.

DISCUSSION: The use of NSAIDs post-operatively has traditionally been limited due to concerns about bleeding risks, however the present study displayed no significant increases in unplanned patient encounters either in the ED or outpatient clinic after ketorolac administration. Our study has several limitations including its retrospective and single-institutional design, difficulties of pain assessment in pediatric population, and possibility of under estimation of unplanned encounters due to limited access to patients' records outside of our institution.

CONCLUSIONS: The use of ketorolac is not associated with an increase in unplanned encounters in children undergoing hypospadias repair. It should be considered a safe agent for perioperative analgesia to decrease opioid utilization. Further studies will evaluate long-term surgical outcomes in children receiving ketorolac after hypospadias repair.

10.1016/j.jpurol.2023.01.014

J Pediatr Urol

36774243

Incidence and resolution of de novo hydronephrosis after pediatric robot-assisted laparoscopic extravesical ureteral reimplantation for primary vesicoureteral reflux.

2022

517.e1-517.e5

08/2022

1873-4898

INTRODUCTION: With the advent of robot-assisted laparoscopic ureteral reimplantation (RALUR) for primary vesicoureteral reflux (VUR), understanding and minimizing its complications continues to be critical. Incidence of de novo hydronephrosis after RALUR could be indicative of an outcome that needs further study or could be a benign transient finding.

OBJECTIVE: In the present study, we hypothesized that the incidence of de novo hydronephrosis after RALUR is low and resolves spontaneously.

METHODS: An IRB-approved prospective single-institutional registry was utilized to identify all patients undergoing RALUR via an extravesical approach from 2012 to 2020. Patients with primary VUR and minimal (Grade I SFU) or no hydronephrosis prior to surgery were included. Patients who had other associated pathology or underwent concomitant procedures were excluded. Preoperative characteristics including VUR and hydronephrosis grades as well as post-operative clinical details and hydronephrosis grades were aggregated and analyzed.

RESULTS: 86/172 (50%) patients (133 ureters), with median reflux grade of 3 (IQR: G2, G3) met the inclusion criteria. Patients underwent RALUR at a median age of 5.7 years (IQR: 3.4, 8.7) with median 36.2 months (IQR: 19.6, 63.6) follow-up. Renal ultrasound at 4-6 weeks post-op showed de novo hydronephrosis in 18 (13.5%) ureters; the complete resolution was seen in 13 ureters (72%) at a median of 4.5 months from surgery. Among the 5 with non-resolved hydronephrosis (SFU G2:4, G3:1), 2 patients (3 ureters) underwent subsequent interventions.

DISCUSSION: The present study evaluating the natural history of de novo hydronephrosis after RALUR-EV performed for primary VUR, is to our knowledge the largest cohort of patients undergoing RALUR that this has been studied in. In our cohort, the incidence of de novo hydronephrosis after RALUR was 13.5%, similar to rates reported in two OUR cohorts, and significantly lower than reported incidence rates of 22-26% in several OUR cohorts, and 30% in a RALUR cohort. In the present cohort, hydronephrosis resolved spontaneously in more than 72% of cases. The median time from surgery until resolution of hydronephrosis was 4.5 (1.6, 10.5) months, which is shorter in comparison to the average time to resolution of 7.6 months, reported by Kim et al. in an earlier study.

CONCLUSIONS: De novo hydronephrosis after RALUR can be followed with serial renal ultrasounds. The majority of de novo hydronephrosis post-RALUR is transient and resolves spontaneously within a year of surgery with a very low re-intervention rate.

10.1016/j.jpurol.2022.04.005

J Pediatr Urol

35654725

Deep learning imaging features derived from kidney ultrasounds predict chronic kidney disease progression in children with posterior urethral valves.

2022

07/2022

1432-198X

BACKGROUND: We sought to use deep learning to extract anatomic features from postnatal kidney ultrasounds and evaluate their performance in predicting the risk and timing of chronic kidney disease (CKD) progression for boys with posterior urethral valves (PUV). We hypothesized that these features would predict CKD progression better than clinical characteristics such as nadir creatinine alone.

METHODS: We performed a retrospective cohort study of boys with PUV treated at two pediatric health systems from 1990 to 2021. Features of kidneys were extracted from initial postnatal kidney ultrasound images using a deep learning model. Three time-to-event prediction models were built using random survival forests. The Imaging Model included deep learning imaging features, the Clinical Model included clinical data, and the Ensemble Model combined imaging features and clinical data. Separate models were built to include time-dependent clinical data that were available at 6 months, 1 year, 3 years, and 5 years.

RESULTS: Two-hundred and twenty-five patients were included in the analysis. All models performed well with C-indices of 0.7 or greater. The Clinical Model outperformed the Imaging Model at all time points with nadir creatinine driving the performance of the Clinical Model. Combining the 6-month Imaging Model (C-index 0.7; 95% confidence interval [CI] 0.6, 0.79) with the 6-month Clinical Model (C-index 0.79; 95% CI 0.71, 0.86) resulted in a 6-month Ensemble Model that performed better (C-index 0.82; 95% CI 0.77, 0.88) than either model alone.

CONCLUSIONS: Deep learning imaging features extracted from initial postnatal kidney ultrasounds may improve early prediction of CKD progression among children with PUV. A higher resolution version of the Graphical abstract is available as Supplementary information.

10.1007/s00467-022-05677-0

Pediatr Nephrol

35867160

Child Abuse Mimic: Avulsion Injury in a Child With Penoscrotal Webbing.

2017

265-267

2017 Apr

1535-1815

<p>Sexual abuse of children is prevalent in today's society. In 2012, approximately 686,000 children (9.2 per 1000) in the United States were determined to be victims of substantiated child abuse and neglect, according to national data compiled by child protective service agencies; victimization rates were highest for children younger than 1 year. Nearly 9.3% of maltreated children were victims of sexual abuse, this finding was reported by US Department of Health and Human Services (http://www.acf.hhs.gov/programs/cb/research-data-technology/statistics-…). Previous research has shown that as many as 1 in 3 girls and 1 in 7 boys will be sexually abused during childhood (Child Abuse Negl. 2003;27:1205-1222). Although sexual abuse seems to be less common in boys than girls, this may be partly due to underdiagnosis and underreporting of sexual abuse in boys (Arch Dis Child. 2007;92:328-331). Clinicians should therefore consider the possibility of sexual abuse when boys present with genital injuries, because failing to recognize and diagnose sexual abuse can pose an ongoing safety risk to a child. However, an erroneous diagnosis of sexual abuse can have equally hazardous repercussions, including removal of a child from their caregivers or prosecution of an innocent individual. A number of medical conditions can mimic child sexual abuse injuries, including anal fissures, failure of midline fusion, perianal streptococcal dermatitis, and straddle injury (J Pediatr Health Care. 2009;23:283-288 and Acta Paediatr. 2011;100:590-593). The following case involves a 5-week-old male infant who presented to the pediatric emergency department with an avulsion injury to his penis concerning for sexual abuse. He was ultimately diagnosed with a relatively rare anatomic variant of the genitalia and determined to have sustained an accidental injury whose appearance mimicked abuse.</p>

10.1097/PEC.0000000000000524

Pediatr Emerg Care

26425931

The natural history of semen parameters in untreated asymptomatic adolescent varicocele patients: A retrospective cohort study.

2016

2016 Oct 24

1873-4898

<p><strong>INTRODUCTION: </strong>Adolescent males with varicoceles present a dilemma for surveillance and treatment. Testicular volumetrics have not been shown to predict SA outcomes. Serial SAs are thus recommended in asymptomatic adolescent males with varicoceles and normal testicular development, but the natural history of semen parameters is unknown.</p>

<p><strong>OBJECTIVE: </strong>To explore the natural history of semen parameters in adolescent boys with a left varicocele under active surveillance.</p>

<p><strong>STUDY DESIGN: </strong>Adolescents with an asymptomatic unilateral left varicocele, Tanner V development, normal testicular volumes, and an initial SA were retrospectively reviewed in a single-institution prospectively followed cohort. Total motile count (TMC) was calculated. A cutoff of TMC Z 20 million was used to dichotomize SA results into "normal" or "poor." Those with poor SA were offered repeat SA. Cumulative probabilities of normal TMC over successive rounds of SA were calculated. Bivariate models were used to explore associations of a second consecutive poor TMC with age and varicocele grade.</p>

<p><strong>RESULTS: </strong>A total of 216 patients provided an initial SA between 1992 and 2015. We excluded 17 for a history of cryptorchidism or incomplete SA data for a final cohort of 199 patients with median follow-up of 3.3 years (interquartile range 1.5-5.6 years). The mean age at initial SA was 17.9 years (range 14.8-21.8 years). One hundred and nine out of 199 had an initial normal TMC. Of the 90 out of 199 with an initially poor TMC, 51 had repeat SA and 24 of the 51 patients improved to normal TMC. Of the 27 patients with two consecutive poor TMCs, 15 had a third SA and five out of 15 improved to normal TMC. Thus, cumulatively, 55%, 67%, and 69% of all patients had a normal TMC after an initial, second, and third SA, respectively. However, fewer patients in each round of SA normalized their TMC (Figure). Neither age nor varicocele grade was associated with a second consecutive poor TMC.</p>

<p><strong>DISCUSSION: </strong>Two-thirds of Tanner V boys with an uncorrected varicocele and normal testicular volumes achieve a normal TMC regardless of varicocele grade or age. Despite Tanner V development, 47% with an initial "poor" SA will improve to normal status without surgery. However, a small subgroup of patients will have persistently poor TMC and thus should be targeted in future research for timely intervention.</p>

<p><strong>CONCLUSION: </strong>Semen parameters improve over time. SA should be followed and repeated at least once in symptomatic Tanner V boys with varicoceles.</p>

10.1016/j.jpurol.2016.09.008

J Pediatr Urol

27815047