Brain volume estimation from post-mortem newborn and fetal MRI.

Area of Science:

• Medical Imaging
• Forensic Pathology
• Pediatric Pathology

Background:

• Post-mortem magnetic resonance imaging (MRI) is a minimally invasive autopsy alternative for fetuses and infants.
• Manual brain segmentation on post-mortem MRI is time-consuming and error-prone, hindering routine clinical use.
• Accurate brain weight estimation is crucial for autopsy but manual methods are impractical.

Purpose of the Study:

• To validate semi-automatic segmentation techniques for accurate post-mortem fetal and neonatal brain volumetric measurements.
• To assess the correlation between MRI-derived brain weights and conventional autopsy results.
• To establish post-mortem MRI as a reliable tool for brain weight estimation in forensic and pediatric autopsies.

Main Methods:

• Post-mortem MR images of 17 newborns and 17 fetuses were segmented using a semi-automatic algorithm and a neonate brain atlas.
• A new atlas was created from averaged segmentations and used for automated atlas-based segmentation.
• Manual segmentation and conventional autopsy weights were used for validation and comparison.

Main Results:

• High Dice overlap coefficients (0.815–0.992) between manual and automatic segmentations across brain regions.
• Excellent agreement between MRI-estimated and conventional autopsy brain weights.
• Low mean absolute differences (5g for fetuses, 20g for newborns) and maximum errors (2% for fetuses, 11% for newborns).

Conclusions:

• Semi-automatic post-mortem MRI segmentation provides accurate brain weight estimations.
• The high correlation with autopsy weights supports post-mortem MRI as a viable alternative to conventional autopsy for brain examination.
• This technique enhances the utility of minimally invasive autopsy in pediatric and forensic cases.