TUBITAK Destekli Projelerden Çıkan Yayınlar Topluluğu
Published January 1, 2025 | Version v1
Conference paper Open

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

Creators

  • 1. Istinye Univ, Artificial Intelligence, Istanbul, Turkiye
  • 2. Istanbul Medipol Univ, Biomed Engn & Bioinformat, Istanbul, Turkiye
  • 3. Istinye Univ, Dept Mech Engn, Istanbul, Turkiye
  • 4. Yildiz Tech Univ, Biomed Engn, Istanbul, Turkiye
  • 5. Istanbul Medipol Univ, Dept Biomed Engn, Istanbul, Turkiye
  • 6. Istinye Univ, Dept Comp Engn, Istanbul, Turkiye
  • 7. Marmara Univ, Bioengn, Istanbul, Turkiye
  • 8. Anadolu Med Ctr, Radiol, Istanbul, Turkiye
  • 9. Medipol Univ, Biomed Engn, Istanbul, Turkiye
  • 10. Istinye Univ, Comp Engn, Istanbul, Turkiye
  • 11. Istinye Univ, Software Engn, Istanbul, Turkiye
  • 12. Sev Amer Coll, Istanbul, Turkiye

Description

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.

Files

bib-ea1eda1a-3da1-4118-a610-70724c011605.txt

Files (305 Bytes)

Name Size Download all
md5:6ef006ff3fa5b66b68ab3266481d0057
305 Bytes Preview Download