Journal Article

·2025

3D U-Net-Based Deep Learning Approach for Pediatric Cardiovascular MRI Segmentation

Muhammad Umair , Mohamad Humam Alzaeim , Şenol Pişkin YTU , Irem Yilmaz YTU , Kevser Banu Köse , J.S. Femilda Josephin , Ömer Faruk Şahin , Fatma Zeynep Güngören , Ayham Pezri , Ahmed Boray ,

Abstract

Accurate segmentation of cardiovascular structures in pediatric patients is essential for clinical evaluation, but variability in imaging data and the labor-intensive nature of manual segmentation remain major challenges. The given paper provides a reproducible deep-learning model intended to counteract these issues. With our deep learning approach, cardiac MRI data is standardized by (i) converting anatomical labels into a four-class mask (involving background, aorta, chambers, pulmonary structures), (ii) aligning images and masks on coordinates using nearest-neighbor resampling, and (iii) automatically cropping the region-of-interest (ROI). To handle class imbalance, we trained a 3D U-Net with the Focal Tversky Loss function, using both public and private clinical datasets. To improve robustness and reduce outlier errors, we applied model ensembling by averaging predictions from independently trained networks. By automating preprocessing and segmentation, our approach provides a pediatric cardiac analysis tool that reduces manual workload and increases reproducibility.

Keywords

Deep learning Segmentation Preprocessor Outlier Robustness (evolution) Pattern recognition (psychology) Workload Artificial intelligence Computer science Computer vision

Subject Areas

Medical Image Segmentation Techniques ·Computer Vision and Pattern Recognition ·Physical Sciences
Advanced Neural Network Applications ·Computer Vision and Pattern Recognition ·Physical Sciences
COVID-19 diagnosis using AI ·Radiology, Nuclear Medicine and Imaging ·Health Sciences