Robust correction of spike noise: application to diffusion tensor imaging.

Area of Science:

• Medical Imaging
• Neuroimaging
• Biophysics

Background:

• Echo-planar imaging (EPI)-based diffusion tensor imaging (DTI) is highly susceptible to spike noise.
• Current spike noise correction methods are often impractical for clinical DTI data as they require the complex MRI signal, which is not typically stored.
• This limitation hinders the quality control and diagnostic utility of DTI scans.

Purpose of the Study:

• To introduce a novel and practical spike noise correction technique for DTI data.
• To develop a method that operates on readily available magnitude images, overcoming limitations of existing approaches.
• To enable automated quality control for DTI datasets.

Main Methods:

• The Outlier Detection De-spiking (ODD) technique involves three stages: detection, localization, and correction of spike noise.
• ODD utilizes automated outlier detection schemes, leveraging data redundancy from multiple diffusion-weighted images (DWIs) within a DTI dataset.
• Mathematical formulations describe spike noise effects on magnitude images for detection and employ normalization-dependent schemes for localization.

Main Results:

• ODD demonstrated accurate performance in removing spike noise across diverse DTI datasets.
• The technique successfully operates on magnitude images, enhancing its applicability in clinical settings.
• The method facilitates automated quality assessment of DTI data.

Conclusions:

• The Outlier Detection De-spiking (ODD) technique provides an effective and practical solution for correcting spike noise in DTI.
• ODD's ability to function with magnitude images makes it suitable for routine clinical use and automated DTI quality control.
• The ODD methodology holds potential for extension to other MRI applications with inherent data redundancy, such as functional MRI and dynamic imaging.