Related Experiment Videos
Automated template-based brain localization and extraction for fetal brain MRI reconstruction
Sébastien Tourbier1, Clemente Velasco-Annis2, Vahid Taimouri2
1Computational Radiology Laboratory (CRL), Department of Radiology, Boston Children's Hospital, and Harvard Medical School, USA; Medical Image Analysis Laboratory (MIAL), Centre d'Imagerie BioMédicale (CIBM), Switzerland; Radiology Department, Lausanne University Hospital Center (CHUV) and University of Lausanne (UNIL), Switzerland.
This study introduces an automated method for fetal brain MRI segmentation using template-based registration, significantly improving reconstruction quality. The new approach automates a previously manual task, achieving high accuracy in fetal brain extraction and analysis.
Area of Science:
- Medical Imaging
- Neuroimaging
- Computational Biology
Background:
- Fetal brain MRI reconstruction requires accurate localization and extraction of brain tissue from low-resolution (LR) images.
- Manual or semi-automatic methods for fetal brain extraction are time-consuming and laborious.
- Existing algorithms often rely solely on brain-relevant voxels, necessitating precise pre-processing steps.
Purpose of the Study:
- To develop an automated method for fetal brain localization and extraction using template-based prior knowledge.
- To integrate this automated method into a comprehensive MRI reconstruction pipeline.
- To enhance the quality and efficiency of fetal brain MRI analysis.
Main Methods:
- A novel automatic brain localization and extraction method employing template-to-slice block matching and deformable slice-to-template registration.
- Integration into a reconstruction pipeline including intensity normalization, inter-slice motion correction, and super-resolution (SR) reconstruction.
- A fusion strategy for projecting LR brain masks into template space for iterative refinement during motion correction.
Main Results:
- The automated algorithm achieved high accuracy in brain mask generation, with an average Dice overlap of 94.5% compared to manual segmentations.
- Slice-to-template registration significantly improved brain extraction performance over global rigid methods.
- Reconstructed image quality was comparable to reference reconstructions using manual brain extraction, as assessed by expert observers.
Conclusions:
- The proposed template-based approach effectively automates fetal brain MRI segmentation and extraction.
- This method offers a significant improvement in processing efficiency and accuracy for fetal brain analysis.
- The pipeline shows promise for automatic fetal brain MRI segmentation and volumetry across a range of gestational ages.