Segmentation of 3-D High-Frequency Ultrasound Images of Human Lymph Nodes Using Graph Cut With Energy Functional...

Area of Science:

• Medical imaging
• Biomedical engineering
• Ultrasound technology

Background:

• Quantitative ultrasound (QUS) shows promise for differentiating metastatic lymph nodes (LNs) from cancer-free LNs.
• Accurate segmentation of LN parenchyma (LNP) and surrounding fat is crucial for QUS analysis, especially for correcting ultrasound attenuation.
• Spatially varying intensity distributions in high-frequency ultrasound images complicate segmentation due to acoustic effects.

Purpose of the Study:

• To develop and validate a novel automatic segmentation method for 3D lymph node ultrasound images.
• To address challenges posed by spatially varying intensity distributions and inhomogeneous acoustic attenuation.
• To improve the accuracy of quantitative ultrasound analysis for metastatic lymph node detection.

Main Methods:

• A novel graph cut (GC) with locally adaptive energy segmentation approach was developed.
• The method was designed to segment lymph node parenchyma (LNP), fat, and surrounding fluid.
• Nested graph cut was previously shown to handle complex boundary cases.

Main Results:

• The proposed GC with locally adaptive energy achieved high segmentation accuracy.
• Dice similarity coefficients of 0.937±0.035 were obtained when compared to expert manual segmentation.
• The method was validated on a dataset of 115 3D LN images from colorectal cancer patients.

Conclusions:

• The novel graph cut segmentation method effectively segments lymph node components in 3D ultrasound images.
• This approach enhances the reliability of quantitative ultrasound analysis for differentiating metastatic lymph nodes.
• The method shows significant potential for improving diagnostic accuracy in cancer patient management.