Automated Society of Fetal Urology (SFU) grading of hydronephrosis on ultrasound imaging using a convolutional neural network.

INTRODUCTION: Grading of hydronephrosis severity on postnatal renal ultrasound guides management decisions in antenatal hydronephrosis (ANH). Multiple systems exist to help standardize hydronephrosis grading, yet poor inter-observer reliability persists. Machine learning methods may provide tools to improve the efficiency and accuracy of hydronephrosis grading. OBJECTIVE: To develop an automated convolutional neural network (CNN) model to classify hydronephrosis on renal ultrasound imaging according to the Society of Fetal Urology (SFU) system as potential clinical adjunct. STUDY DESIGN: A cross-sectional, single-institution cohort of postnatal renal ultrasounds with radiologist SFU grading from pediatric patients with and without hydronephrosis of stable severity was obtained. Imaging labels were used to automatedly select sagittal and transverse grey-scale renal images from all available studies from each patient. A VGG16 pre-trained ImageNet CNN model analyzed these preprocessed images. Three-fold stratified cross-validation was used to build and evaluate the model that was used to classify renal ultrasounds on a per patient basis into five classes based on the SFU system (normal, SFU I, SFU II, SFU III, or SFU IV). These predictions were compared to radiologist grading. Confusion matrices evaluated model performance. Gradient class activation mapping demonstrated imaging features driving model predictions. RESULTS: We identified 710 patients with 4659 postnatal renal ultrasound series. Per radiologist grading, 183 were normal, 157 were SFU I, 132 were SFU II, 100 were SFU III, and 138 were SFU IV. The machine learning model predicted hydronephrosis grade with 82.0% (95% CI: 75-83%) overall accuracy and classified 97.6% (95% CI: 95-98%) of the patients correctly or within one grade of the radiologist grade. The model classified 92.3% (95% CI: 86-95%) normal, 73.2% (95% CI: 69-76%) SFU I, 73.5% (95% CI: 67-75%) SFU II, 79.0% (95% CI: 73-82%) SFU III, and 88.4% (95% CI: 85-92%) SFU IV patients accurately. Gradient class activation mapping demonstrated that the ultrasound appearance of the renal collecting system drove the model's predictions. DISCUSSION: The CNN-based model classified hydronephrosis on renal ultrasounds automatically and accurately based on the expected imaging features in the SFU system. Compared to prior studies, the model functioned more automatically with greater accuracy. Limitations include the retrospective, relatively small cohort, and averaging across multiple imaging studies per patient. CONCLUSIONS: An automated CNN-based system classified hydronephrosis on renal ultrasounds according to the SFU system with promising accuracy based on appropriate imaging features. These findings suggest a possible adjunctive role for machine learning systems in the grading of ANH.

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

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.

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

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.

Bilateral ureteral reimplantation at complete primary repair of exstrophy: Post-operative outcomes.

BACKGROUND AND STUDY OBJECTIVE: The value of bilateral ureteral reimplant (BUR) at the time of complete primary repair of bladder exstrophy (CPRE) has been suggested, however, outcomes are poorly characterized in current medical literature. We hypothesize that BUR at time of CPRE will decrease the rate of recurrent pyelonephritis, post-operative vesicoureteral reflux (VUR), and the need for subsequent ureteral surgery. STUDY DESIGN: We analyzed 64 consecutive patients with a diagnosis of classic bladder exstrophy (BE) who underwent CPRE at three institutions from 2013 to 2019.15 patients underwent cephalotrigonal BUR-CPRE and 49 patients underwent CPRE alone. Our primary outcome was >1 episode of pyelonephritis as documented in the medical record. Secondary outcomes were persistent vesicoureteral reflux (VUR), with a sub-analysis of number of refluxing renal units and presence of dilating VUR, and the need for subsequent ureteral surgery. Descriptive statistics in addition to standard, two tailed univariate statistics, were used to compare the groups where appropriate. RESULTS: BUR-CPRE was associated with a significant decrease in the rates of post-operative VUR, number of refluxing renal units, and need for subsequent ureteral surgery (p = 0.002, p = 0.001, and p = 0.048 respectively). There was a reduction in the rates of recurrent pyelonephritis and dilating reflux in patients undergoing BUR-CPRE, though it did not reach significance. Female gender was significantly associated with recurrent pyelonephritis regardless of BUR-CPRE status (p = 0.005). There were no reports of distal ureteral obstruction or other complications following BUR-CPRE. The mean post-operative follow up for the BUR-CPRE group was 46.33 (10.26) months vs. 53.76 (26.05) months for CPRE (p = 0.11). DISCUSSION: Recurrent pyelonephritis following bladder closure in patients with BE is a well-documented surgical complication, with centers performing CPRE reporting rates of post-operative pyelonephritis from 22 to 50%. Our series demonstrates similar efficacy of BUR-CPRE compared to other contemporary series and provides additional detail about need for subsequent ureteral surgeries and increased long term follow-up of these complex patients. Limitations of the study include male predominance of the cohort and lack of randomization of BUR-CPRE. CONCLUSIONS: BUR-CPRE decreases postoperative VUR and the need for additional ureteral surgery in select BE patients; it should be considered when technically feasible. While results continue to suggest a trend toward decreased recurrent pyelonephritis and dilating reflux, further longitudinal follow-up in our cohort will be needed.

Prostate Magnetic Resonance Imaging Provides Limited Incremental Value Over the Memorial Sloan Kettering Cancer Center Preradical Prostatectomy Nomogram.

OBJECTIVE: To examine the incremental value of prostate magnetic resonance imaging (MRI) when used in combination with the currently available preoperative risk stratification tool, the Memorial Sloan Kettering Cancer Center (MSKCC) preradical prostatectomy nomogram. MATERIALS AND METHODS: We reviewed our institutional database of prostate MRI performed before radical prostatectomy between December 2014 and March 2016 (n = 236). We generated a logistic regression model based on observed final pathology results and the MSKCC nomogram predictions for organ-confined disease, extracapsular extension (ECE), seminal vesicle invasion, and lymph node involvement (LNI) ("MSKCC only"). We then generated a combined regression model incorporating both the MSKCC nomogram prediction with the degree of prostate MRI suspicion ("MSKCC + MRI"). Receiver operating characteristic curves were generated, and the area under the curves (AUCs) were compared. RESULTS: When independently examining the MSKCC nomogram predicted risk and the degree of prostate MRI suspicion, MRI was a predictor for ECE (odds ratio 2.8, P <.01) and LNI (odds ratio 5.6, P = .01). When examining the "MSKCC + MRI" and "MSKCC only" models, the incremental benefit in risk discrimination from the combined model ("MSKCC + MRI") was not significant for organ-confined disease, ECE, seminal vesicle invasion, or LNI (ΔAUC +0.03, P = .10; ΔAUC +0.03, P = .08; ΔAUC 0.63, P = .63; ΔAUC +0.04, P = .42; respectively). CONCLUSION: A combined model with prostate MRI and the MSKCC nomogram provides no additional risk discrimination over the MSKCC nomogram-based model alone. Evaluation of prostate MRI as a predictive tool should be performed in combination with, not independent of, these clinical risk stratification models.

Magnetic Resonance Imaging Provides Added Value to the Prostate Cancer Prevention Trial Risk Calculator for Patients With Estimated Risk of High-grade Prostate Cancer Less Than or Equal to 10.

OBJECTIVE: To determine the added value of prostate magnetic resonance imaging (MRI) to the Prostate Cancer Prevention Trial risk calculator. METHODS: Between January 2012 and December 2015, 339 patients underwent prostate MRI prior to biopsy at our institution. MRI was considered positive if there was at least 1 Prostate Imaging Reporting and Data System 4 or 5 MRI suspicious region. Logistic regression was used to develop 2 models: biopsy outcome as a function of the (1) Prostate Cancer Prevention Trial risk calculator alone and (2) combined with MRI findings. RESULTS: When including all patients, the Prostate Cancer Prevention Trial with and without MRI models performed similarly (area under the curve [AUC] = 0.74 and 0.78, P = .06). When restricting the cohort to patients with estimated risk of high-grade (Gleason ≥7) prostate cancer ≤10%, the model with MRI outperformed the Prostate Cancer Prevention Trial alone model (AUC = 0.69 and 0.60, P = .01). Within this cohort of patients, there was no significant difference in discrimination between models for those with previous negative biopsy (AUC = 0.61 vs 0.63, P = .76), whereas there was a significant improvement in discrimination with the MRI model for biopsy-naïve patients (AUC = 0.72 vs 0.60, P = .01). CONCLUSION: The use of prostate MRI in addition to the Prostate Cancer Prevention Trial risk calculator provides a significant improvement in clinical risk discrimination for patients with estimated risk of high-grade (Gleason ≥7) prostate cancer ≤10%. Prebiopsy prostate MRI should be strongly considered for these patients.

Presence of Magnetic Resonance Imaging Suspicious Lesion Predicts Gleason 7 or Greater Prostate Cancer in Biopsy-Naive Patients.

OBJECTIVE: To compare the relative value of magnetic resonance imaging (MRI) in biopsy-naive patients to those with previous negative biopsy. Although MRI-targeted biopsy has been studied in several major prostate cancer (PCa) cohorts (biopsy naive, previous negative biopsy, and active surveillance), the relative benefit in these cohorts has not been established. METHODS: We retrospectively reviewed biopsy-naive (n = 45) and previous negative biopsy (n = 55) patients who underwent prostate MRI prior to biopsy at our institution. Patients with an MRI suspicious region (MSR) underwent MRI-targeted biopsy as well as a systematic template biopsy, whereas those without MSR underwent only the template biopsy. All biopsies were performed with the TargetScan (Envisioneering, Pittsburgh, PA) biopsy system. MRI targeting was performed with cognitive guidance. RESULTS: On multivariate logistic regression, the presence of an MSR was the only statistically significant and independent predictor of Gleason ≥ 7 PCa on biopsy for biopsy-naive men (odds ratio [OR] 40.2, P = .01). For men with previous negative biopsy, the presence of MSR was not a predictor of Gleason ≥ 7 PCa on biopsy (OR 4.35, P = .16), whereas PSA density > 0.15 ng/mL(2) was a significant and independent predictor (OR 66.2, P < .01). CONCLUSION: Prostate MRI should be considered prior to biopsy in all patients presenting with clinical suspicion for PCa, as presence of a MSR will help guide prebiopsy counseling and provide an opportunity for MRI targeting during biopsy.