Optimization of breast lesion segmentation in texture feature space approach

Area of Science:

• Medical imaging analysis
• Computational pathology
• Image segmentation techniques

Background:

• Accurate breast lesion boundary detection is crucial for diagnosis and treatment planning.
• Existing methods may lack precision in segmenting complex or subtle lesion features.
• Integrating texture analysis with active contour models offers a promising avenue for improved segmentation.

Purpose of the Study:

• To develop and evaluate a semi-automatic method for breast lesion boundary detection.
• To combine snake evolution techniques with statistical texture information for enhanced segmentation accuracy.
• To identify the most relevant image features for effective lesion segmentation.

Main Methods:

• Utilized snake evolution techniques for image segmentation.
• Incorporated statistical texture information, including first-order textural features and n×n masks.
• Developed an efficient image energy function based on image and textural features.
• Evaluated segmentation results using area error rate.
• Assessed image features qualitatively using contrast-to-noise ratio and fractal dimension analysis.

Main Results:

• The proposed method demonstrates effective semi-automatic detection of breast lesion boundaries.
• Standard deviation, skewness, and entropy were identified as the most relevant image features for segmentation.
• The segmentation approach showed promising results in terms of accuracy, evaluated by area error rate.

Conclusions:

• The combination of snake evolution and statistical texture analysis provides an efficient and accurate method for breast lesion segmentation.
• The identified key image features (standard deviation, skewness, entropy) are critical for improving segmentation performance.
• This technique holds potential for enhancing computer-aided diagnosis systems in mammography and other breast imaging modalities.