Related Experiment Video
Updated: Apr 2, 2026

Automated Joint Space Detection Improves Bone Segmentation Accuracy
Published on: November 28, 2025
KidMesh: Computational Mesh Reconstruction for Pediatric Congenital Hydronephrosis Using Deep Neural Networks.
This study introduces KidMesh, a deep learning method that automatically creates 3D models of pediatric congenital hydronephrosis from MRU scans. KidMesh enables faster, more accurate functional assessments for improved clinical diagnosis and treatment planning.
Area of Science:
- Medical Imaging
- Computational Biology
- Pediatric Urology
Background:
- Pediatric congenital hydronephrosis (CH) is a common urinary tract obstruction.
- Magnetic resonance urography (MRU) visualizes CH, but current segmentation methods focus on morphology, requiring complex post-processing for functional analysis.
- Existing methods necessitate extensive post-processing to convert segmented CH regions into mesh representations for urodynamic simulations.
Purpose of the Study:
- To develop an end-to-end deep learning method, KidMesh, for direct reconstruction of CH meshes from MRU images.
- To enable functional assessments, such as urodynamic simulations, by generating accurate mesh-level representations.
- To overcome the limitations of current segmentation techniques that require complex post-processing.
Main Methods:
- KidMesh utilizes deep neural networks to extract feature maps from MRU images.
- Feature maps are converted into feature vertices via grid sampling, which then deform a template mesh to generate patient-specific CH meshes.
- A novel training schema was developed to train KidMesh without requiring precise mesh-level annotations, addressing challenges posed by sparsely sampled MRU data.
Main Results:
- KidMesh reconstructs CH meshes rapidly, averaging 0.4 seconds per case.
- The method achieves comparable performance to conventional approaches without post-processing.
- Reconstructed meshes demonstrate high accuracy, with a Dice score of 0.86 against manual segmentations and minimal vertex error distances (3.7% > 3.2mm, 0.2% > 6.4mm).
Conclusions:
- KidMesh offers an efficient and automated solution for reconstructing CH meshes directly from MRU.
- The generated meshes facilitate urodynamic simulations, providing valuable functional insights for clinical practice.
- This approach enhances the diagnostic and prognostic capabilities for pediatric congenital hydronephrosis.
More Related Videos
09:10Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
08:03Midface Hypoplasia and Cranial Base Morphology in Syndromic Craniosynostosis: A Comparative Analysis Study Using a Predictive Regression Model
Published on: November 4, 2025