Three-dimensional region-based segmentation for breast tumors on sonography.

Area of Science:

• Medical Imaging
• Radiology
• Computational Biology

Background:

• Breast tumor shape and size on sonography are crucial for differentiating malignant and benign types.
• Accurate diagnosis from sonograms is challenging due to inherent noise and tissue texture, relying heavily on clinician expertise.
• Manual 3D breast tumor contouring is laborious and complex, hindering efficient clinical workflow.

Purpose of the Study:

• To develop an efficient method for automatic 3D contour detection of breast tumors in 3D sonography.
• To provide an automated contouring solution that mimics manual sketching for improved diagnostic accuracy.
• To reduce the time and complexity associated with manual breast tumor contour delineation.

Main Methods:

• Image preprocessing using voxel nearest neighbor, Wiener, and unsharp filters to enhance contrast and reduce noise.
• Application of a 3D region-growing algorithm for initial breast tumor contour extraction.
• Postprocessing steps to refine the segmented contour and minimize shadow regions.

Main Results:

• The proposed 3D segmentation method demonstrated robust contouring performance on breast sonograms.
• Computer simulations showed contours generated by the automated method closely resemble those from manual contouring.
• The technique successfully reduced the time required for precise breast tumor contour sketching.

Conclusions:

• The developed automatic 3D contour detection method is efficient and accurate for breast sonography.
• This automated approach can significantly aid clinicians in diagnosing breast tumors by providing reliable contours.
• The method offers a valuable tool for enhancing diagnostic accuracy and streamlining the clinical workflow in breast imaging.