Combinatorial active contour bilateral filter for ultrasound image segmentation.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Image Segmentation

Background:

• Computer-aided diagnosis (CAD) in radiological ultrasound (US) imaging is increasingly vital for cancer detection.
• Accurate segmentation of cancerous lesions in US images is critical for clinical recommendations but challenging due to noise and low contrast.

Purpose of the Study:

• To develop and evaluate a novel framework for segmenting lesions in breast and thyroid ultrasound images.
• To enhance the accuracy of lesion segmentation for improved computer-aided diagnosis systems.

Main Methods:

• A combinatorial framework utilizing a bilateral filter (BF) for image smoothing and edge preservation.
• A hybrid region-edge-based active contour (AC) model applied globally-to-locally for lesion area capture.
• Validation on 258 breast and thyroid US images against manual ground truths.

Main Results:

• The proposed framework achieved high performance in lesion segmentation, quantified by the Dice coefficient.
• The inclusion of the bilateral filter significantly improved the segmentation framework's performance.
• Quantitative evaluation demonstrated the effectiveness of the proposed segmentation method.

Conclusions:

• The developed segmentation framework shows high performance and potential for practical application in CAD radiological US systems.
• The method offers a robust solution for overcoming segmentation challenges in US imaging.
• This work contributes to advancing the accuracy and reliability of computer-aided diagnosis in radiology.