Get Started
Home
Topics
Search
Library
Research questionHow can we reliably segment multiple pediatric brain structures in ultra-low-field MRI when annotations disagree?At 0.064 T, weak boundaries and partially visible small structures make anatomical labels uncertain. High-field-derived masks can also be locally misregistered relative to anatomy visible in low-field scans, creating conflicting supervision.
AI
Computer Vision
Health
Image & Video Processing
Machine Learning
Research Paper
Latest papersRecent research connected to this question, newest first.Asymmetric Paired-Annotation Learning for Multi-Structure ULF Pediatric Brain MRI SegmentationThe report studies paired high-field-derived and low-field-edited annotations for 0.064 T pediatric brain MRI using an nnU-Net-based asymmetric supervision strategy. Evidence comes from a 16-case development split; hidden-test and external-cohort performance remains unassessed.research paper · Sep 2, 2026
Related questions
How can automated quality control reliably grade artifact severity in ultra-low-field neonatal brain MRI with practical inference costs?How can brain MRI super-resolution preserve continuous tissue mixtures at partial-volume tissue transitions?How can semi-supervised medical image segmentation prevent appearance variation from corrupting structural cues and pseudo-labels?How can 3D brain MRI inpainting reconstruct healthy tissue in pathological regions without changing observed anatomy?