Information measures-based intensity standardization of MRI

Area of Science:

• Medical Imaging
• Image Processing
• Computational Anatomy

Background:

• Scan-to-scan intensity variation in medical images complicates intensity-based processing.
• Existing intensity standardization methods have limitations, especially with high-intensity regions.

Purpose of the Study:

• To develop a novel, flexible intensity standardization technique for magnetic resonance (MR) images.
• To improve the performance of intensity-based image processing methods like segmentation and registration.

Main Methods:

• Utilized information measures, specifically Kullback-Leibler (KL) divergence for pairwise image standardization.
• Employed minimum entropy as a group probability similarity measure for group-wise standardization.
• Optimized intensity mappings using Powell's search on cost functions.

Main Results:

• The proposed method offers greater flexibility in mapping high-intensity regions without hard limits.
• Demonstrated improved performance in non-rigid registration and feature map-based segmentation of MR brain images.
• Achieved unbiased standardization for group-wise MR images.

Conclusions:

• The novel intensity standardization technique effectively addresses limitations of current methods.
• This approach enhances the accuracy and reliability of downstream image analysis tasks.
• The method shows promise for clinical applications requiring robust MR image standardization.