A modified fuzzy C-means method for segmenting MR images using non-local information.

Area of Science:

• Medical Imaging
• Image Processing
• Computational Biology

Background:

• Magnetic Resonance (MR) images are crucial for therapeutic applications like image-guided radiotherapy (IGRT).
• Low contrast and noise in MR images pose significant challenges for accurate image segmentation.
• Effective segmentation is vital for precise treatment planning and delivery in IGRT.

Purpose of the Study:

• To develop a robust method for segmenting MR images corrupted by Gaussian noise.
• To improve the accuracy and reliability of MR image segmentation using a modified Fuzzy C-Means (FCM) algorithm.

Main Methods:

• A modified FCM algorithm was developed, incorporating non-local pixel information through Hausdorff distance.
• The algorithm's membership and objective functions were adjusted to enhance segmentation performance.
• Segmentations were compared using varying weights for the Hausdorff distance to assess its impact.

Main Results:

• The proposed algorithm demonstrated superior boundary resolution and robustness against Gaussian noise in synthetic and MR images.
• Tests confirmed the method's capability in accurately capturing the centroid of target regions, as shown with a sample tumor MR image.
• The modified FCM algorithm effectively segmented noisy and low-contrast MR images.

Conclusions:

• The enhanced FCM algorithm, utilizing neighboring pixel information, is effective for segmenting blurry MR images.
• This method holds significant potential for applications in segmenting motion MR images within image-guided radiotherapy (IGRT).
• The developed technique offers a promising solution for improving image segmentation accuracy in challenging clinical scenarios.