This repository provides the official implementation of a deep learning pipeline for the automatic classification of Posterior Urethral Valves (PUV) from Voiding CystoUrethroGram (VCUG) images, based on EfficientNet architectures.
Posterior Urethral Valves (PUV) are a rare congenital anomaly affecting the male urethra. Early and accurate diagnosis is crucial to prevent serious complications, including renal damage. This project aims to support the diagnostic process by exploiting Convolutional Neural Networks (CNN) to identify PUV cases directly from imaging data.
- Contributes to the early diagnosis of PUV, which is essential to avoid long-term renal complications.
- Provides a reproducible framework that can be adapted or extended to other urological conditions.
- Promotes collaboration between clinicians and AI researchers, bridging the gap between computer vision and healthcare.
The dataset used in this study originates from a real-world acquisition campaign conducted within the activity of the SUD4VUP project, coordinated by the University of Campania “Luigi Vanvitelli”.
The data collection involved the Nephrology and Urology Units at "Luigi Vanvitelli" University Hospital, AORN Santobono-Pausilipon in Naples, and IRCCS Ca’ Granda in Milan.
The dataset includes routine VCUG acquisitions from male pediatric patients, each labeled to indicate the presence or absence of Posterior Urethral Valves (PUV).
Due to privacy regulations and clinical data protection protocols, the dataset is not publicly available. However, preprocessing scripts and data handling instructions are included to support reproducibility using compatible datasets.
Researchers interested in collaboration or data access for institutional research purposes may contact the project coordinators.
The development of this repository involved contributions from a multidisciplinary team of researchers, including:
- Ciro Russo - PostDoctoral Researcher, University of Cassino and Lazio Meridionale
- Gaetano Settembre - PhD Student in Computer Science and Mathematics, University of Bari Aldo Moro
- Grazia Gargano - PhD Student in Computer Science and Mathematics, University of Bari Aldo Moro
- Maria Stella de Biase - Researcher, University of Campania "Luigi Vanvitelli"
- Roberta De Fazio - PostDoctoral Fellowship, University of Campania "Luigi Vanvitelli"
This work was developed through a collaboration between:
- University of Cassino and Lazio Meridionale
- University of Bari "Aldo Moro"
- University of Campania "Luigi Vanvitelli" (coordinator of the SUP4VUP project and data provider)
If you find the project codes useful for your research, please consider citing
@InProceedings{PUVClassificationICIAP25,
title = {{AI in Pediatric Urology: Deep Learning-based Approach supporting Posterior Urethral Valves Diagnosis on VCUG Imaging}},
ISBN = {9783032113818},
ISSN = {1611-3349},
editor = {Rodotà, Emanuele and Galasso, Fabio and Masi, Iacopo},
DOI = {10.1007/978-3-032-11381-8_12},
booktitle = {Image Analysis and Processing - ICIAP 2025 Workshops},
publisher = {Springer Nature Switzerland},
address = {Cham},
author = {Russo, Ciro and Settembre, Gaetano and Gargano, Grazia and {de Biase}, Maria Stella and {De Fazio}, Roberta},
year = {2026},
pages = {138–149}
}