An improved graph cut segmentation method for cervical lymph nodes on sonograms and its relationship with node's...

Area of Science:

• Medical Imaging
• Image Segmentation
• Ultrasound Technology

Background:

• Accurate segmentation of cervical lymph nodes in ultrasound (sonogram) images is crucial for diagnosis.
• Traditional graph cut methods may struggle with precise segmentation, impacting subsequent measurements.
• The short axis to long axis ratio (S/L) is a key metric for evaluating lymph node characteristics.

Purpose of the Study:

• To develop and evaluate a modified graph cut algorithm for segmenting cervical lymph nodes.
• To investigate the impact of an elliptical shape constraint on segmentation accuracy.
• To assess the algorithm's performance in measuring the short axis to long axis ratio (S/L) using relative ultimate measurement accuracy (RUMA).

Main Methods:

• A modified graph cut algorithm incorporating an elliptical shape prior was developed.
• The algorithm was applied to segment cervical lymph nodes in 60 sonogram images.
• Performance was evaluated by comparing segmentation results and S/L ratio measurements against manual segmentation and traditional graph cut methods using RUMA.

Main Results:

• The proposed modified graph cut algorithm successfully segmented all 60 tested sonograms.
• The traditional graph cut method failed to segment all images under identical conditions.
• The mean RUMA for S/L measurement using the developed method was comparable to manual segmentation.
• The elliptical shape prior significantly improved graph cut performance for lymph node segmentation.

Conclusions:

• The modified graph cut algorithm with an elliptical shape constraint offers improved accuracy for cervical lymph node segmentation in ultrasound.
• The method meets the required accuracy standards for S/L ratio measurements.
• Incorporating shape priors is an effective strategy to enhance graph cut-based image segmentation.