EVALUATION OF SELECTED 2-DIMENSIONAL SEGMENTATION TECHNIQUES FOR COMPUTED-TOMOGRAPHY QUANTITATION OF LYMPH-NODES

RATIONALE AND OBJECTIVES. AS contrast agents that selectively target n ormal lymph nodes are undergoing development and evaluation, it has be come important to accurately and reproducibly determine nodal boundari es to study. the agents and determine such values as lymph node area o r mean nodal contrast concentration. This study was performed to evalu ate the accuracy of different two-dimensional computer segmentation me thods, tested on acrylic phantoms constructed to imitate the appearanc e of lymph nodes surrounded bg fat. METHODS. Five segmentation techniq ues (manual tracing, semiautomatic local criteria threshold selection, Sobel/watershed technique, interactive deformable contour algorithm a nd thresholding) were evaluated using phantoms, Subsequently, the firs t three methods were applied to the images of enhanced lymph nodes in rabbits. RESULTS. Minimum errors in phantom area measurement (<5%) and interoperator variation (<5%) were seen with the Sobel/watershed tech nique and the interactive deformable contour algorithm, These two tech niques were significantly better than thresholding and semiautomated t hresholding based on local properties. CONCLUSION. Methods based on So bel edge detection offer more objective tools than thresholding method s for segmenting objects similar to lymph nodes in computed tomography images, Both methods, Sobel/watershed and interactive deformable cont our algorithm, are fast and have simple user interfaces.