Mammographic images segmentation based on chaotic map clustering algorithm.

Area of Science:

• Medical Imaging
• Computational Intelligence
• Image Processing

Background:

• Investigates a novel clustering approach for segmenting digital mammographic images.
• Utilizes a chaotic map clustering algorithm to partition mammograms into meaningful regions.

Purpose of the Study:

• To assess the applicability of chaotic map clustering for mammographic image segmentation.
• To evaluate the algorithm's effectiveness in identifying lesions and micro-calcifications.

Main Methods:

• Image pixels are grouped into subsets based on selected features.
• A mutual coupling strength is introduced between feature space points.
• Chaotic dynamics simulate map evolution, leading to image partitioning.

Main Results:

• Achieved a high recognition rate (94%) for segmenting small mass lesions.
• Successfully reproduced shapes of micro-calcification regions in approximately 2/3 of cases.
• Demonstrated lower effectiveness in identifying larger mass lesions.

Conclusions:

• Chaotic map clustering is not ideal as a standalone mammographic segmentation method.
• Combining this algorithm with other techniques enhances segmentation performance.
• The approach can provide valuable information for subsequent analysis, such as ROI classification.