Preterm neonatal diffusion processing using detection and replacement of outliers prior to resampling.

Area of Science:

• Medical Imaging
• Neuroscience
• Biomedical Engineering

Background:

• Subject motion and brain pulsation degrade diffusion-weighted MRI (dMRI) quality, causing signal drop-outs and misalignment.
• Unsedated neonates are particularly susceptible to motion artifacts, impairing diffusion tensor estimation.
• Existing methods like retrospective registration and robust estimators have limitations when combined, as resampling can obscure outliers.

Purpose of the Study:

• To present a novel method for removing outliers in dMRI data prior to resampling.
• To account for image misalignment while removing motion and intensity outliers.
• To improve the accuracy of diffusion tensor estimation in neonates, especially those with high motion.

Main Methods:

• Developed a method to identify and remove outlier voxels before image resampling.
• Integrated misalignment correction into the outlier removal process.
• Compared the proposed method against existing processing pipelines using simulated data and real dMRI scans from unsedated preterm neonates.

Main Results:

• The proposed method effectively removes outliers while preserving image integrity.
• Demonstrated improved diffusion tensor estimation compared to standard processing pipelines.
• Showed particular benefits in processing dMRI data from neonates exhibiting high levels of motion.

Conclusions:

• The novel outlier removal technique enhances dMRI scan quality in challenging populations like neonates.
• This approach mitigates the negative impact of motion and pulsation artifacts on diffusion tensor estimation.
• The method offers a significant advantage for analyzing dMRI data from infants with substantial movement.