Segmentation of solid nodules in ultrasonographic breast image based on wavelet transform

Area of Science:

• Medical Imaging
• Image Processing
• Radiology

Background:

• Accurate segmentation of solid nodules in breast ultrasonography is crucial for diagnosis.
• Texture analysis provides valuable information for differentiating nodule types.

Purpose of the Study:

• To develop and present an accurate algorithm for segmenting solid nodules in breast ultrasound images.
• To utilize texture features for classification and segmentation of breast lesions.

Main Methods:

• Employed 1-level 2-dimensional Discrete Wavelet Transform (DWT) for texture feature extraction.
• Implemented texture classification based on extracted features.
• Developed an automatic threshold determination method for texture classification, informed by empirical data.

Main Results:

• Successfully segmented solid nodule regions from classified textures.
• The algorithm was applied to a dataset of 284 malignant and 300 benign solid nodules.
• Visualizations of the segmentation results were generated.

Conclusions:

• The presented algorithm demonstrates accurate segmentation of solid breast nodules.
• Texture analysis using DWT is effective for nodule classification and segmentation in ultrasonography.
• The automated thresholding approach enhances the robustness of the segmentation process.