Nasopharyngeal carcinoma segmentation using a region growing technique

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Oncology

Background:

• Accurate segmentation of nasopharyngeal tumors in CT images is crucial for diagnosis and treatment planning.
• Traditional seeded region growing (SRG) methods are sensitive to initial seed point selection and image intensity variations.

Purpose of the Study:

• To develop and evaluate a novel image segmentation technique for identifying nasopharyngeal tumor regions in CT images.
• To improve upon the limitations of the traditional SRG approach by incorporating Self-Organizing Maps (SOM).

Main Methods:

• A modified SRG technique was developed, integrating SOM for initial tumor region localization and mode intensity for seed point determination.
• CT images from nasopharyngeal carcinoma (NPC) patients were analyzed, with tumor regions delineated by expert radiologists to establish ground truth.
• The proposed method was applied to localized CT image groups (Group II and III) based on anatomical location.

Main Results:

• The modified SRG technique demonstrated superior performance in segmenting nasopharyngeal tumor regions compared to the traditional SRG approach.
• Average corresponding ratios (CRs) of 0.67 and 0.69 were achieved for Group II and III, respectively.
• Average segmentation accuracy (PMs) reached 78.17% for Group II and 82.47% for Group III, indicating high efficiency.

Conclusions:

• The proposed modified SRG technique, utilizing SOM and mode intensity, is an effective and efficient method for nasopharyngeal tumor segmentation in CT images.
• This approach offers improved accuracy and robustness over traditional SRG methods, reducing sensitivity to initial seed placement.
• The findings suggest significant potential for clinical application in the diagnosis and management of nasopharyngeal carcinoma.