Joint brain parametric T1-map segmentation and RF inhomogeneity calibration

Area of Science:

• Medical Imaging
• Computational Neuroscience
• Image Processing

Background:

• Accurate brain magnetic resonance imaging (MRI) segmentation is crucial for neurological studies.
• Existing segmentation models may require enhancements for multi-tissue classification and artifact correction.
• Parametric T(1)-Maps and T(1)-weighted images offer valuable insights into brain tissue properties.

Purpose of the Study:

• To develop a systematic procedure for segmenting brain MRI data using a constrained Mumford-Shah model.
• To enable accurate 3-tissue segmentation of T(1)-Map and T(1)-weighted images in 2D and 3D.
• To introduce a novel method for joint segmentation and radiofrequency (RF) inhomogeneity calibration (JSRIC).

Main Methods:

• A constrained Mumford-Shah segmentation model incorporating an information-theoretic perspective.
• Integration of a tuning weight for probabilistic segmentation and 3-tissue classification.
• Development of the Joint Segmentation and RF Inhomogeneity Calibration (JSRIC) method for T(1)-Map calibration.

Main Results:

• The proposed segmentation method successfully performs 3-tissue segmentation.
• JSRIC effectively calibrates RF inhomogeneity, generating accurate T(1)-Maps by rectifying flip angles.
• The methods were validated on human subjects and public brain MRI databases (BrainWeb, IBSR).

Conclusions:

• The information-theoretic segmentation approach provides a robust method for brain MRI analysis.
• JSRIC significantly improves the accuracy of T(1)-Map generation by addressing RF inhomogeneity.
• This work offers enhanced tools for quantitative analysis of brain structure and function from MRI data.